2 * linux/fs/ext4/super.c
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/inode.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
19 #include <linux/module.h>
20 #include <linux/string.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
45 #include "ext4_jbd2.h"
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
53 struct proc_dir_entry
*ext4_proc_root
;
54 static struct kset
*ext4_kset
;
56 static int ext4_load_journal(struct super_block
*, struct ext4_super_block
*,
57 unsigned long journal_devnum
);
58 static int ext4_commit_super(struct super_block
*sb
, int sync
);
59 static void ext4_mark_recovery_complete(struct super_block
*sb
,
60 struct ext4_super_block
*es
);
61 static void ext4_clear_journal_err(struct super_block
*sb
,
62 struct ext4_super_block
*es
);
63 static int ext4_sync_fs(struct super_block
*sb
, int wait
);
64 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
66 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
);
67 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
);
68 static int ext4_unfreeze(struct super_block
*sb
);
69 static void ext4_write_super(struct super_block
*sb
);
70 static int ext4_freeze(struct super_block
*sb
);
71 static int ext4_get_sb(struct file_system_type
*fs_type
, int flags
,
72 const char *dev_name
, void *data
, struct vfsmount
*mnt
);
74 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
75 static struct file_system_type ext3_fs_type
= {
78 .get_sb
= ext4_get_sb
,
79 .kill_sb
= kill_block_super
,
80 .fs_flags
= FS_REQUIRES_DEV
,
82 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
84 #define IS_EXT3_SB(sb) (0)
87 ext4_fsblk_t
ext4_block_bitmap(struct super_block
*sb
,
88 struct ext4_group_desc
*bg
)
90 return le32_to_cpu(bg
->bg_block_bitmap_lo
) |
91 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
92 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_block_bitmap_hi
) << 32 : 0);
95 ext4_fsblk_t
ext4_inode_bitmap(struct super_block
*sb
,
96 struct ext4_group_desc
*bg
)
98 return le32_to_cpu(bg
->bg_inode_bitmap_lo
) |
99 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
100 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_bitmap_hi
) << 32 : 0);
103 ext4_fsblk_t
ext4_inode_table(struct super_block
*sb
,
104 struct ext4_group_desc
*bg
)
106 return le32_to_cpu(bg
->bg_inode_table_lo
) |
107 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
108 (ext4_fsblk_t
)le32_to_cpu(bg
->bg_inode_table_hi
) << 32 : 0);
111 __u32
ext4_free_blks_count(struct super_block
*sb
,
112 struct ext4_group_desc
*bg
)
114 return le16_to_cpu(bg
->bg_free_blocks_count_lo
) |
115 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
116 (__u32
)le16_to_cpu(bg
->bg_free_blocks_count_hi
) << 16 : 0);
119 __u32
ext4_free_inodes_count(struct super_block
*sb
,
120 struct ext4_group_desc
*bg
)
122 return le16_to_cpu(bg
->bg_free_inodes_count_lo
) |
123 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
124 (__u32
)le16_to_cpu(bg
->bg_free_inodes_count_hi
) << 16 : 0);
127 __u32
ext4_used_dirs_count(struct super_block
*sb
,
128 struct ext4_group_desc
*bg
)
130 return le16_to_cpu(bg
->bg_used_dirs_count_lo
) |
131 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
132 (__u32
)le16_to_cpu(bg
->bg_used_dirs_count_hi
) << 16 : 0);
135 __u32
ext4_itable_unused_count(struct super_block
*sb
,
136 struct ext4_group_desc
*bg
)
138 return le16_to_cpu(bg
->bg_itable_unused_lo
) |
139 (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
?
140 (__u32
)le16_to_cpu(bg
->bg_itable_unused_hi
) << 16 : 0);
143 void ext4_block_bitmap_set(struct super_block
*sb
,
144 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
146 bg
->bg_block_bitmap_lo
= cpu_to_le32((u32
)blk
);
147 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
148 bg
->bg_block_bitmap_hi
= cpu_to_le32(blk
>> 32);
151 void ext4_inode_bitmap_set(struct super_block
*sb
,
152 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
154 bg
->bg_inode_bitmap_lo
= cpu_to_le32((u32
)blk
);
155 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
156 bg
->bg_inode_bitmap_hi
= cpu_to_le32(blk
>> 32);
159 void ext4_inode_table_set(struct super_block
*sb
,
160 struct ext4_group_desc
*bg
, ext4_fsblk_t blk
)
162 bg
->bg_inode_table_lo
= cpu_to_le32((u32
)blk
);
163 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
164 bg
->bg_inode_table_hi
= cpu_to_le32(blk
>> 32);
167 void ext4_free_blks_set(struct super_block
*sb
,
168 struct ext4_group_desc
*bg
, __u32 count
)
170 bg
->bg_free_blocks_count_lo
= cpu_to_le16((__u16
)count
);
171 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
172 bg
->bg_free_blocks_count_hi
= cpu_to_le16(count
>> 16);
175 void ext4_free_inodes_set(struct super_block
*sb
,
176 struct ext4_group_desc
*bg
, __u32 count
)
178 bg
->bg_free_inodes_count_lo
= cpu_to_le16((__u16
)count
);
179 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
180 bg
->bg_free_inodes_count_hi
= cpu_to_le16(count
>> 16);
183 void ext4_used_dirs_set(struct super_block
*sb
,
184 struct ext4_group_desc
*bg
, __u32 count
)
186 bg
->bg_used_dirs_count_lo
= cpu_to_le16((__u16
)count
);
187 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
188 bg
->bg_used_dirs_count_hi
= cpu_to_le16(count
>> 16);
191 void ext4_itable_unused_set(struct super_block
*sb
,
192 struct ext4_group_desc
*bg
, __u32 count
)
194 bg
->bg_itable_unused_lo
= cpu_to_le16((__u16
)count
);
195 if (EXT4_DESC_SIZE(sb
) >= EXT4_MIN_DESC_SIZE_64BIT
)
196 bg
->bg_itable_unused_hi
= cpu_to_le16(count
>> 16);
200 /* Just increment the non-pointer handle value */
201 static handle_t
*ext4_get_nojournal(void)
203 handle_t
*handle
= current
->journal_info
;
204 unsigned long ref_cnt
= (unsigned long)handle
;
206 BUG_ON(ref_cnt
>= EXT4_NOJOURNAL_MAX_REF_COUNT
);
209 handle
= (handle_t
*)ref_cnt
;
211 current
->journal_info
= handle
;
216 /* Decrement the non-pointer handle value */
217 static void ext4_put_nojournal(handle_t
*handle
)
219 unsigned long ref_cnt
= (unsigned long)handle
;
221 BUG_ON(ref_cnt
== 0);
224 handle
= (handle_t
*)ref_cnt
;
226 current
->journal_info
= handle
;
230 * Wrappers for jbd2_journal_start/end.
232 * The only special thing we need to do here is to make sure that all
233 * journal_end calls result in the superblock being marked dirty, so
234 * that sync() will call the filesystem's write_super callback if
237 handle_t
*ext4_journal_start_sb(struct super_block
*sb
, int nblocks
)
241 if (sb
->s_flags
& MS_RDONLY
)
242 return ERR_PTR(-EROFS
);
244 vfs_check_frozen(sb
, SB_FREEZE_WRITE
);
245 /* Special case here: if the journal has aborted behind our
246 * backs (eg. EIO in the commit thread), then we still need to
247 * take the FS itself readonly cleanly. */
248 journal
= EXT4_SB(sb
)->s_journal
;
250 if (is_journal_aborted(journal
)) {
251 ext4_abort(sb
, "Detected aborted journal");
252 return ERR_PTR(-EROFS
);
254 return jbd2_journal_start(journal
, nblocks
);
256 return ext4_get_nojournal();
260 * The only special thing we need to do here is to make sure that all
261 * jbd2_journal_stop calls result in the superblock being marked dirty, so
262 * that sync() will call the filesystem's write_super callback if
265 int __ext4_journal_stop(const char *where
, unsigned int line
, handle_t
*handle
)
267 struct super_block
*sb
;
271 if (!ext4_handle_valid(handle
)) {
272 ext4_put_nojournal(handle
);
275 sb
= handle
->h_transaction
->t_journal
->j_private
;
277 rc
= jbd2_journal_stop(handle
);
282 __ext4_std_error(sb
, where
, line
, err
);
286 void ext4_journal_abort_handle(const char *caller
, unsigned int line
,
287 const char *err_fn
, struct buffer_head
*bh
,
288 handle_t
*handle
, int err
)
291 const char *errstr
= ext4_decode_error(NULL
, err
, nbuf
);
293 BUG_ON(!ext4_handle_valid(handle
));
296 BUFFER_TRACE(bh
, "abort");
301 if (is_handle_aborted(handle
))
304 printk(KERN_ERR
"%s:%d: aborting transaction: %s in %s\n",
305 caller
, line
, errstr
, err_fn
);
307 jbd2_journal_abort_handle(handle
);
310 static void __save_error_info(struct super_block
*sb
, const char *func
,
313 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
315 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
316 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
317 es
->s_last_error_time
= cpu_to_le32(get_seconds());
318 strncpy(es
->s_last_error_func
, func
, sizeof(es
->s_last_error_func
));
319 es
->s_last_error_line
= cpu_to_le32(line
);
320 if (!es
->s_first_error_time
) {
321 es
->s_first_error_time
= es
->s_last_error_time
;
322 strncpy(es
->s_first_error_func
, func
,
323 sizeof(es
->s_first_error_func
));
324 es
->s_first_error_line
= cpu_to_le32(line
);
325 es
->s_first_error_ino
= es
->s_last_error_ino
;
326 es
->s_first_error_block
= es
->s_last_error_block
;
329 * Start the daily error reporting function if it hasn't been
332 if (!es
->s_error_count
)
333 mod_timer(&EXT4_SB(sb
)->s_err_report
, jiffies
+ 24*60*60*HZ
);
334 es
->s_error_count
= cpu_to_le32(le32_to_cpu(es
->s_error_count
) + 1);
337 static void save_error_info(struct super_block
*sb
, const char *func
,
340 __save_error_info(sb
, func
, line
);
341 ext4_commit_super(sb
, 1);
345 /* Deal with the reporting of failure conditions on a filesystem such as
346 * inconsistencies detected or read IO failures.
348 * On ext2, we can store the error state of the filesystem in the
349 * superblock. That is not possible on ext4, because we may have other
350 * write ordering constraints on the superblock which prevent us from
351 * writing it out straight away; and given that the journal is about to
352 * be aborted, we can't rely on the current, or future, transactions to
353 * write out the superblock safely.
355 * We'll just use the jbd2_journal_abort() error code to record an error in
356 * the journal instead. On recovery, the journal will complain about
357 * that error until we've noted it down and cleared it.
360 static void ext4_handle_error(struct super_block
*sb
)
362 if (sb
->s_flags
& MS_RDONLY
)
365 if (!test_opt(sb
, ERRORS_CONT
)) {
366 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
368 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
370 jbd2_journal_abort(journal
, -EIO
);
372 if (test_opt(sb
, ERRORS_RO
)) {
373 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
374 sb
->s_flags
|= MS_RDONLY
;
376 if (test_opt(sb
, ERRORS_PANIC
))
377 panic("EXT4-fs (device %s): panic forced after error\n",
381 void __ext4_error(struct super_block
*sb
, const char *function
,
382 unsigned int line
, const char *fmt
, ...)
387 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: comm %s: ",
388 sb
->s_id
, function
, line
, current
->comm
);
393 ext4_handle_error(sb
);
396 void ext4_error_inode(struct inode
*inode
, const char *function
,
397 unsigned int line
, ext4_fsblk_t block
,
398 const char *fmt
, ...)
401 struct ext4_super_block
*es
= EXT4_SB(inode
->i_sb
)->s_es
;
403 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
404 es
->s_last_error_block
= cpu_to_le64(block
);
405 save_error_info(inode
->i_sb
, function
, line
);
407 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: inode #%lu: ",
408 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
);
410 printk("block %llu: ", block
);
411 printk("comm %s: ", current
->comm
);
416 ext4_handle_error(inode
->i_sb
);
419 void ext4_error_file(struct file
*file
, const char *function
,
420 unsigned int line
, const char *fmt
, ...)
423 struct ext4_super_block
*es
;
424 struct inode
*inode
= file
->f_dentry
->d_inode
;
425 char pathname
[80], *path
;
427 es
= EXT4_SB(inode
->i_sb
)->s_es
;
428 es
->s_last_error_ino
= cpu_to_le32(inode
->i_ino
);
429 save_error_info(inode
->i_sb
, function
, line
);
431 path
= d_path(&(file
->f_path
), pathname
, sizeof(pathname
));
435 "EXT4-fs error (device %s): %s:%d: inode #%lu "
436 "(comm %s path %s): ",
437 inode
->i_sb
->s_id
, function
, line
, inode
->i_ino
,
438 current
->comm
, path
);
443 ext4_handle_error(inode
->i_sb
);
446 static const char *ext4_decode_error(struct super_block
*sb
, int errno
,
453 errstr
= "IO failure";
456 errstr
= "Out of memory";
459 if (!sb
|| (EXT4_SB(sb
)->s_journal
&&
460 EXT4_SB(sb
)->s_journal
->j_flags
& JBD2_ABORT
))
461 errstr
= "Journal has aborted";
463 errstr
= "Readonly filesystem";
466 /* If the caller passed in an extra buffer for unknown
467 * errors, textualise them now. Else we just return
470 /* Check for truncated error codes... */
471 if (snprintf(nbuf
, 16, "error %d", -errno
) >= 0)
480 /* __ext4_std_error decodes expected errors from journaling functions
481 * automatically and invokes the appropriate error response. */
483 void __ext4_std_error(struct super_block
*sb
, const char *function
,
484 unsigned int line
, int errno
)
489 /* Special case: if the error is EROFS, and we're not already
490 * inside a transaction, then there's really no point in logging
492 if (errno
== -EROFS
&& journal_current_handle() == NULL
&&
493 (sb
->s_flags
& MS_RDONLY
))
496 errstr
= ext4_decode_error(sb
, errno
, nbuf
);
497 printk(KERN_CRIT
"EXT4-fs error (device %s) in %s:%d: %s\n",
498 sb
->s_id
, function
, line
, errstr
);
499 save_error_info(sb
, function
, line
);
501 ext4_handle_error(sb
);
505 * ext4_abort is a much stronger failure handler than ext4_error. The
506 * abort function may be used to deal with unrecoverable failures such
507 * as journal IO errors or ENOMEM at a critical moment in log management.
509 * We unconditionally force the filesystem into an ABORT|READONLY state,
510 * unless the error response on the fs has been set to panic in which
511 * case we take the easy way out and panic immediately.
514 void __ext4_abort(struct super_block
*sb
, const char *function
,
515 unsigned int line
, const char *fmt
, ...)
519 save_error_info(sb
, function
, line
);
521 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: ", sb
->s_id
,
527 if ((sb
->s_flags
& MS_RDONLY
) == 0) {
528 ext4_msg(sb
, KERN_CRIT
, "Remounting filesystem read-only");
529 sb
->s_flags
|= MS_RDONLY
;
530 EXT4_SB(sb
)->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
531 if (EXT4_SB(sb
)->s_journal
)
532 jbd2_journal_abort(EXT4_SB(sb
)->s_journal
, -EIO
);
533 save_error_info(sb
, function
, line
);
535 if (test_opt(sb
, ERRORS_PANIC
))
536 panic("EXT4-fs panic from previous error\n");
539 void ext4_msg (struct super_block
* sb
, const char *prefix
,
540 const char *fmt
, ...)
545 printk("%sEXT4-fs (%s): ", prefix
, sb
->s_id
);
551 void __ext4_warning(struct super_block
*sb
, const char *function
,
552 unsigned int line
, const char *fmt
, ...)
557 printk(KERN_WARNING
"EXT4-fs warning (device %s): %s:%d: ",
558 sb
->s_id
, function
, line
);
564 void __ext4_grp_locked_error(const char *function
, unsigned int line
,
565 struct super_block
*sb
, ext4_group_t grp
,
566 unsigned long ino
, ext4_fsblk_t block
,
567 const char *fmt
, ...)
572 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
574 es
->s_last_error_ino
= cpu_to_le32(ino
);
575 es
->s_last_error_block
= cpu_to_le64(block
);
576 __save_error_info(sb
, function
, line
);
578 printk(KERN_CRIT
"EXT4-fs error (device %s): %s:%d: group %u",
579 sb
->s_id
, function
, line
, grp
);
581 printk("inode %lu: ", ino
);
583 printk("block %llu:", (unsigned long long) block
);
588 if (test_opt(sb
, ERRORS_CONT
)) {
589 ext4_commit_super(sb
, 0);
593 ext4_unlock_group(sb
, grp
);
594 ext4_handle_error(sb
);
596 * We only get here in the ERRORS_RO case; relocking the group
597 * may be dangerous, but nothing bad will happen since the
598 * filesystem will have already been marked read/only and the
599 * journal has been aborted. We return 1 as a hint to callers
600 * who might what to use the return value from
601 * ext4_grp_locked_error() to distinguish beween the
602 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
603 * aggressively from the ext4 function in question, with a
604 * more appropriate error code.
606 ext4_lock_group(sb
, grp
);
610 void ext4_update_dynamic_rev(struct super_block
*sb
)
612 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
614 if (le32_to_cpu(es
->s_rev_level
) > EXT4_GOOD_OLD_REV
)
618 "updating to rev %d because of new feature flag, "
619 "running e2fsck is recommended",
622 es
->s_first_ino
= cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO
);
623 es
->s_inode_size
= cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE
);
624 es
->s_rev_level
= cpu_to_le32(EXT4_DYNAMIC_REV
);
625 /* leave es->s_feature_*compat flags alone */
626 /* es->s_uuid will be set by e2fsck if empty */
629 * The rest of the superblock fields should be zero, and if not it
630 * means they are likely already in use, so leave them alone. We
631 * can leave it up to e2fsck to clean up any inconsistencies there.
636 * Open the external journal device
638 static struct block_device
*ext4_blkdev_get(dev_t dev
, struct super_block
*sb
)
640 struct block_device
*bdev
;
641 char b
[BDEVNAME_SIZE
];
643 bdev
= open_by_devnum(dev
, FMODE_READ
|FMODE_WRITE
);
649 ext4_msg(sb
, KERN_ERR
, "failed to open journal device %s: %ld",
650 __bdevname(dev
, b
), PTR_ERR(bdev
));
655 * Release the journal device
657 static int ext4_blkdev_put(struct block_device
*bdev
)
660 return blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
663 static int ext4_blkdev_remove(struct ext4_sb_info
*sbi
)
665 struct block_device
*bdev
;
668 bdev
= sbi
->journal_bdev
;
670 ret
= ext4_blkdev_put(bdev
);
671 sbi
->journal_bdev
= NULL
;
676 static inline struct inode
*orphan_list_entry(struct list_head
*l
)
678 return &list_entry(l
, struct ext4_inode_info
, i_orphan
)->vfs_inode
;
681 static void dump_orphan_list(struct super_block
*sb
, struct ext4_sb_info
*sbi
)
685 ext4_msg(sb
, KERN_ERR
, "sb orphan head is %d",
686 le32_to_cpu(sbi
->s_es
->s_last_orphan
));
688 printk(KERN_ERR
"sb_info orphan list:\n");
689 list_for_each(l
, &sbi
->s_orphan
) {
690 struct inode
*inode
= orphan_list_entry(l
);
692 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
693 inode
->i_sb
->s_id
, inode
->i_ino
, inode
,
694 inode
->i_mode
, inode
->i_nlink
,
699 static void ext4_put_super(struct super_block
*sb
)
701 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
702 struct ext4_super_block
*es
= sbi
->s_es
;
705 dquot_disable(sb
, -1, DQUOT_USAGE_ENABLED
| DQUOT_LIMITS_ENABLED
);
707 flush_workqueue(sbi
->dio_unwritten_wq
);
708 destroy_workqueue(sbi
->dio_unwritten_wq
);
713 ext4_commit_super(sb
, 1);
715 if (sbi
->s_journal
) {
716 err
= jbd2_journal_destroy(sbi
->s_journal
);
717 sbi
->s_journal
= NULL
;
719 ext4_abort(sb
, "Couldn't clean up the journal");
722 ext4_release_system_zone(sb
);
724 ext4_ext_release(sb
);
725 ext4_xattr_put_super(sb
);
727 if (!(sb
->s_flags
& MS_RDONLY
)) {
728 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
729 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
730 ext4_commit_super(sb
, 1);
733 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
735 kobject_del(&sbi
->s_kobj
);
737 for (i
= 0; i
< sbi
->s_gdb_count
; i
++)
738 brelse(sbi
->s_group_desc
[i
]);
739 kfree(sbi
->s_group_desc
);
740 if (is_vmalloc_addr(sbi
->s_flex_groups
))
741 vfree(sbi
->s_flex_groups
);
743 kfree(sbi
->s_flex_groups
);
744 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
745 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
746 percpu_counter_destroy(&sbi
->s_dirs_counter
);
747 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
750 for (i
= 0; i
< MAXQUOTAS
; i
++)
751 kfree(sbi
->s_qf_names
[i
]);
754 /* Debugging code just in case the in-memory inode orphan list
755 * isn't empty. The on-disk one can be non-empty if we've
756 * detected an error and taken the fs readonly, but the
757 * in-memory list had better be clean by this point. */
758 if (!list_empty(&sbi
->s_orphan
))
759 dump_orphan_list(sb
, sbi
);
760 J_ASSERT(list_empty(&sbi
->s_orphan
));
762 invalidate_bdev(sb
->s_bdev
);
763 if (sbi
->journal_bdev
&& sbi
->journal_bdev
!= sb
->s_bdev
) {
765 * Invalidate the journal device's buffers. We don't want them
766 * floating about in memory - the physical journal device may
767 * hotswapped, and it breaks the `ro-after' testing code.
769 sync_blockdev(sbi
->journal_bdev
);
770 invalidate_bdev(sbi
->journal_bdev
);
771 ext4_blkdev_remove(sbi
);
773 sb
->s_fs_info
= NULL
;
775 * Now that we are completely done shutting down the
776 * superblock, we need to actually destroy the kobject.
780 kobject_put(&sbi
->s_kobj
);
781 wait_for_completion(&sbi
->s_kobj_unregister
);
782 kfree(sbi
->s_blockgroup_lock
);
786 static struct kmem_cache
*ext4_inode_cachep
;
789 * Called inside transaction, so use GFP_NOFS
791 static struct inode
*ext4_alloc_inode(struct super_block
*sb
)
793 struct ext4_inode_info
*ei
;
795 ei
= kmem_cache_alloc(ext4_inode_cachep
, GFP_NOFS
);
799 ei
->vfs_inode
.i_version
= 1;
800 ei
->vfs_inode
.i_data
.writeback_index
= 0;
801 memset(&ei
->i_cached_extent
, 0, sizeof(struct ext4_ext_cache
));
802 INIT_LIST_HEAD(&ei
->i_prealloc_list
);
803 spin_lock_init(&ei
->i_prealloc_lock
);
805 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
806 * therefore it can be null here. Don't check it, just initialize
809 jbd2_journal_init_jbd_inode(&ei
->jinode
, &ei
->vfs_inode
);
810 ei
->i_reserved_data_blocks
= 0;
811 ei
->i_reserved_meta_blocks
= 0;
812 ei
->i_allocated_meta_blocks
= 0;
813 ei
->i_da_metadata_calc_len
= 0;
814 ei
->i_delalloc_reserved_flag
= 0;
815 spin_lock_init(&(ei
->i_block_reservation_lock
));
817 ei
->i_reserved_quota
= 0;
819 INIT_LIST_HEAD(&ei
->i_completed_io_list
);
820 spin_lock_init(&ei
->i_completed_io_lock
);
821 ei
->cur_aio_dio
= NULL
;
823 ei
->i_datasync_tid
= 0;
825 return &ei
->vfs_inode
;
828 static void ext4_destroy_inode(struct inode
*inode
)
830 if (!list_empty(&(EXT4_I(inode
)->i_orphan
))) {
831 ext4_msg(inode
->i_sb
, KERN_ERR
,
832 "Inode %lu (%p): orphan list check failed!",
833 inode
->i_ino
, EXT4_I(inode
));
834 print_hex_dump(KERN_INFO
, "", DUMP_PREFIX_ADDRESS
, 16, 4,
835 EXT4_I(inode
), sizeof(struct ext4_inode_info
),
839 kmem_cache_free(ext4_inode_cachep
, EXT4_I(inode
));
842 static void init_once(void *foo
)
844 struct ext4_inode_info
*ei
= (struct ext4_inode_info
*) foo
;
846 INIT_LIST_HEAD(&ei
->i_orphan
);
847 #ifdef CONFIG_EXT4_FS_XATTR
848 init_rwsem(&ei
->xattr_sem
);
850 init_rwsem(&ei
->i_data_sem
);
851 inode_init_once(&ei
->vfs_inode
);
854 static int init_inodecache(void)
856 ext4_inode_cachep
= kmem_cache_create("ext4_inode_cache",
857 sizeof(struct ext4_inode_info
),
858 0, (SLAB_RECLAIM_ACCOUNT
|
861 if (ext4_inode_cachep
== NULL
)
866 static void destroy_inodecache(void)
868 kmem_cache_destroy(ext4_inode_cachep
);
871 static void ext4_clear_inode(struct inode
*inode
)
874 ext4_discard_preallocations(inode
);
875 if (EXT4_JOURNAL(inode
))
876 jbd2_journal_release_jbd_inode(EXT4_SB(inode
->i_sb
)->s_journal
,
877 &EXT4_I(inode
)->jinode
);
880 static inline void ext4_show_quota_options(struct seq_file
*seq
,
881 struct super_block
*sb
)
883 #if defined(CONFIG_QUOTA)
884 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
886 if (sbi
->s_jquota_fmt
) {
889 switch (sbi
->s_jquota_fmt
) {
900 seq_printf(seq
, ",jqfmt=%s", fmtname
);
903 if (sbi
->s_qf_names
[USRQUOTA
])
904 seq_printf(seq
, ",usrjquota=%s", sbi
->s_qf_names
[USRQUOTA
]);
906 if (sbi
->s_qf_names
[GRPQUOTA
])
907 seq_printf(seq
, ",grpjquota=%s", sbi
->s_qf_names
[GRPQUOTA
]);
909 if (test_opt(sb
, USRQUOTA
))
910 seq_puts(seq
, ",usrquota");
912 if (test_opt(sb
, GRPQUOTA
))
913 seq_puts(seq
, ",grpquota");
919 * - it's set to a non-default value OR
920 * - if the per-sb default is different from the global default
922 static int ext4_show_options(struct seq_file
*seq
, struct vfsmount
*vfs
)
925 unsigned long def_mount_opts
;
926 struct super_block
*sb
= vfs
->mnt_sb
;
927 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
928 struct ext4_super_block
*es
= sbi
->s_es
;
930 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
931 def_errors
= le16_to_cpu(es
->s_errors
);
933 if (sbi
->s_sb_block
!= 1)
934 seq_printf(seq
, ",sb=%llu", sbi
->s_sb_block
);
935 if (test_opt(sb
, MINIX_DF
))
936 seq_puts(seq
, ",minixdf");
937 if (test_opt(sb
, GRPID
) && !(def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
938 seq_puts(seq
, ",grpid");
939 if (!test_opt(sb
, GRPID
) && (def_mount_opts
& EXT4_DEFM_BSDGROUPS
))
940 seq_puts(seq
, ",nogrpid");
941 if (sbi
->s_resuid
!= EXT4_DEF_RESUID
||
942 le16_to_cpu(es
->s_def_resuid
) != EXT4_DEF_RESUID
) {
943 seq_printf(seq
, ",resuid=%u", sbi
->s_resuid
);
945 if (sbi
->s_resgid
!= EXT4_DEF_RESGID
||
946 le16_to_cpu(es
->s_def_resgid
) != EXT4_DEF_RESGID
) {
947 seq_printf(seq
, ",resgid=%u", sbi
->s_resgid
);
949 if (test_opt(sb
, ERRORS_RO
)) {
950 if (def_errors
== EXT4_ERRORS_PANIC
||
951 def_errors
== EXT4_ERRORS_CONTINUE
) {
952 seq_puts(seq
, ",errors=remount-ro");
955 if (test_opt(sb
, ERRORS_CONT
) && def_errors
!= EXT4_ERRORS_CONTINUE
)
956 seq_puts(seq
, ",errors=continue");
957 if (test_opt(sb
, ERRORS_PANIC
) && def_errors
!= EXT4_ERRORS_PANIC
)
958 seq_puts(seq
, ",errors=panic");
959 if (test_opt(sb
, NO_UID32
) && !(def_mount_opts
& EXT4_DEFM_UID16
))
960 seq_puts(seq
, ",nouid32");
961 if (test_opt(sb
, DEBUG
) && !(def_mount_opts
& EXT4_DEFM_DEBUG
))
962 seq_puts(seq
, ",debug");
963 if (test_opt(sb
, OLDALLOC
))
964 seq_puts(seq
, ",oldalloc");
965 #ifdef CONFIG_EXT4_FS_XATTR
966 if (test_opt(sb
, XATTR_USER
) &&
967 !(def_mount_opts
& EXT4_DEFM_XATTR_USER
))
968 seq_puts(seq
, ",user_xattr");
969 if (!test_opt(sb
, XATTR_USER
) &&
970 (def_mount_opts
& EXT4_DEFM_XATTR_USER
)) {
971 seq_puts(seq
, ",nouser_xattr");
974 #ifdef CONFIG_EXT4_FS_POSIX_ACL
975 if (test_opt(sb
, POSIX_ACL
) && !(def_mount_opts
& EXT4_DEFM_ACL
))
976 seq_puts(seq
, ",acl");
977 if (!test_opt(sb
, POSIX_ACL
) && (def_mount_opts
& EXT4_DEFM_ACL
))
978 seq_puts(seq
, ",noacl");
980 if (sbi
->s_commit_interval
!= JBD2_DEFAULT_MAX_COMMIT_AGE
*HZ
) {
981 seq_printf(seq
, ",commit=%u",
982 (unsigned) (sbi
->s_commit_interval
/ HZ
));
984 if (sbi
->s_min_batch_time
!= EXT4_DEF_MIN_BATCH_TIME
) {
985 seq_printf(seq
, ",min_batch_time=%u",
986 (unsigned) sbi
->s_min_batch_time
);
988 if (sbi
->s_max_batch_time
!= EXT4_DEF_MAX_BATCH_TIME
) {
989 seq_printf(seq
, ",max_batch_time=%u",
990 (unsigned) sbi
->s_min_batch_time
);
994 * We're changing the default of barrier mount option, so
995 * let's always display its mount state so it's clear what its
998 seq_puts(seq
, ",barrier=");
999 seq_puts(seq
, test_opt(sb
, BARRIER
) ? "1" : "0");
1000 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
))
1001 seq_puts(seq
, ",journal_async_commit");
1002 else if (test_opt(sb
, JOURNAL_CHECKSUM
))
1003 seq_puts(seq
, ",journal_checksum");
1004 if (test_opt(sb
, I_VERSION
))
1005 seq_puts(seq
, ",i_version");
1006 if (!test_opt(sb
, DELALLOC
))
1007 seq_puts(seq
, ",nodelalloc");
1011 seq_printf(seq
, ",stripe=%lu", sbi
->s_stripe
);
1013 * journal mode get enabled in different ways
1014 * So just print the value even if we didn't specify it
1016 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
1017 seq_puts(seq
, ",data=journal");
1018 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
1019 seq_puts(seq
, ",data=ordered");
1020 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
1021 seq_puts(seq
, ",data=writeback");
1023 if (sbi
->s_inode_readahead_blks
!= EXT4_DEF_INODE_READAHEAD_BLKS
)
1024 seq_printf(seq
, ",inode_readahead_blks=%u",
1025 sbi
->s_inode_readahead_blks
);
1027 if (test_opt(sb
, DATA_ERR_ABORT
))
1028 seq_puts(seq
, ",data_err=abort");
1030 if (test_opt(sb
, NO_AUTO_DA_ALLOC
))
1031 seq_puts(seq
, ",noauto_da_alloc");
1033 if (test_opt(sb
, DISCARD
))
1034 seq_puts(seq
, ",discard");
1036 if (test_opt(sb
, NOLOAD
))
1037 seq_puts(seq
, ",norecovery");
1039 if (test_opt(sb
, DIOREAD_NOLOCK
))
1040 seq_puts(seq
, ",dioread_nolock");
1042 ext4_show_quota_options(seq
, sb
);
1047 static struct inode
*ext4_nfs_get_inode(struct super_block
*sb
,
1048 u64 ino
, u32 generation
)
1050 struct inode
*inode
;
1052 if (ino
< EXT4_FIRST_INO(sb
) && ino
!= EXT4_ROOT_INO
)
1053 return ERR_PTR(-ESTALE
);
1054 if (ino
> le32_to_cpu(EXT4_SB(sb
)->s_es
->s_inodes_count
))
1055 return ERR_PTR(-ESTALE
);
1057 /* iget isn't really right if the inode is currently unallocated!!
1059 * ext4_read_inode will return a bad_inode if the inode had been
1060 * deleted, so we should be safe.
1062 * Currently we don't know the generation for parent directory, so
1063 * a generation of 0 means "accept any"
1065 inode
= ext4_iget(sb
, ino
);
1067 return ERR_CAST(inode
);
1068 if (generation
&& inode
->i_generation
!= generation
) {
1070 return ERR_PTR(-ESTALE
);
1076 static struct dentry
*ext4_fh_to_dentry(struct super_block
*sb
, struct fid
*fid
,
1077 int fh_len
, int fh_type
)
1079 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
1080 ext4_nfs_get_inode
);
1083 static struct dentry
*ext4_fh_to_parent(struct super_block
*sb
, struct fid
*fid
,
1084 int fh_len
, int fh_type
)
1086 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
1087 ext4_nfs_get_inode
);
1091 * Try to release metadata pages (indirect blocks, directories) which are
1092 * mapped via the block device. Since these pages could have journal heads
1093 * which would prevent try_to_free_buffers() from freeing them, we must use
1094 * jbd2 layer's try_to_free_buffers() function to release them.
1096 static int bdev_try_to_free_page(struct super_block
*sb
, struct page
*page
,
1099 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
1101 WARN_ON(PageChecked(page
));
1102 if (!page_has_buffers(page
))
1105 return jbd2_journal_try_to_free_buffers(journal
, page
,
1106 wait
& ~__GFP_WAIT
);
1107 return try_to_free_buffers(page
);
1111 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1112 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1114 static int ext4_write_dquot(struct dquot
*dquot
);
1115 static int ext4_acquire_dquot(struct dquot
*dquot
);
1116 static int ext4_release_dquot(struct dquot
*dquot
);
1117 static int ext4_mark_dquot_dirty(struct dquot
*dquot
);
1118 static int ext4_write_info(struct super_block
*sb
, int type
);
1119 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
1121 static int ext4_quota_on_mount(struct super_block
*sb
, int type
);
1122 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
1123 size_t len
, loff_t off
);
1124 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
1125 const char *data
, size_t len
, loff_t off
);
1127 static const struct dquot_operations ext4_quota_operations
= {
1129 .get_reserved_space
= ext4_get_reserved_space
,
1131 .write_dquot
= ext4_write_dquot
,
1132 .acquire_dquot
= ext4_acquire_dquot
,
1133 .release_dquot
= ext4_release_dquot
,
1134 .mark_dirty
= ext4_mark_dquot_dirty
,
1135 .write_info
= ext4_write_info
,
1136 .alloc_dquot
= dquot_alloc
,
1137 .destroy_dquot
= dquot_destroy
,
1140 static const struct quotactl_ops ext4_qctl_operations
= {
1141 .quota_on
= ext4_quota_on
,
1142 .quota_off
= dquot_quota_off
,
1143 .quota_sync
= dquot_quota_sync
,
1144 .get_info
= dquot_get_dqinfo
,
1145 .set_info
= dquot_set_dqinfo
,
1146 .get_dqblk
= dquot_get_dqblk
,
1147 .set_dqblk
= dquot_set_dqblk
1151 static const struct super_operations ext4_sops
= {
1152 .alloc_inode
= ext4_alloc_inode
,
1153 .destroy_inode
= ext4_destroy_inode
,
1154 .write_inode
= ext4_write_inode
,
1155 .dirty_inode
= ext4_dirty_inode
,
1156 .delete_inode
= ext4_delete_inode
,
1157 .put_super
= ext4_put_super
,
1158 .sync_fs
= ext4_sync_fs
,
1159 .freeze_fs
= ext4_freeze
,
1160 .unfreeze_fs
= ext4_unfreeze
,
1161 .statfs
= ext4_statfs
,
1162 .remount_fs
= ext4_remount
,
1163 .clear_inode
= ext4_clear_inode
,
1164 .show_options
= ext4_show_options
,
1166 .quota_read
= ext4_quota_read
,
1167 .quota_write
= ext4_quota_write
,
1169 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1172 static const struct super_operations ext4_nojournal_sops
= {
1173 .alloc_inode
= ext4_alloc_inode
,
1174 .destroy_inode
= ext4_destroy_inode
,
1175 .write_inode
= ext4_write_inode
,
1176 .dirty_inode
= ext4_dirty_inode
,
1177 .delete_inode
= ext4_delete_inode
,
1178 .write_super
= ext4_write_super
,
1179 .put_super
= ext4_put_super
,
1180 .statfs
= ext4_statfs
,
1181 .remount_fs
= ext4_remount
,
1182 .clear_inode
= ext4_clear_inode
,
1183 .show_options
= ext4_show_options
,
1185 .quota_read
= ext4_quota_read
,
1186 .quota_write
= ext4_quota_write
,
1188 .bdev_try_to_free_page
= bdev_try_to_free_page
,
1191 static const struct export_operations ext4_export_ops
= {
1192 .fh_to_dentry
= ext4_fh_to_dentry
,
1193 .fh_to_parent
= ext4_fh_to_parent
,
1194 .get_parent
= ext4_get_parent
,
1198 Opt_bsd_df
, Opt_minix_df
, Opt_grpid
, Opt_nogrpid
,
1199 Opt_resgid
, Opt_resuid
, Opt_sb
, Opt_err_cont
, Opt_err_panic
, Opt_err_ro
,
1200 Opt_nouid32
, Opt_debug
, Opt_oldalloc
, Opt_orlov
,
1201 Opt_user_xattr
, Opt_nouser_xattr
, Opt_acl
, Opt_noacl
,
1202 Opt_auto_da_alloc
, Opt_noauto_da_alloc
, Opt_noload
, Opt_nobh
, Opt_bh
,
1203 Opt_commit
, Opt_min_batch_time
, Opt_max_batch_time
,
1204 Opt_journal_update
, Opt_journal_dev
,
1205 Opt_journal_checksum
, Opt_journal_async_commit
,
1206 Opt_abort
, Opt_data_journal
, Opt_data_ordered
, Opt_data_writeback
,
1207 Opt_data_err_abort
, Opt_data_err_ignore
,
1208 Opt_usrjquota
, Opt_grpjquota
, Opt_offusrjquota
, Opt_offgrpjquota
,
1209 Opt_jqfmt_vfsold
, Opt_jqfmt_vfsv0
, Opt_jqfmt_vfsv1
, Opt_quota
,
1210 Opt_noquota
, Opt_ignore
, Opt_barrier
, Opt_nobarrier
, Opt_err
,
1211 Opt_resize
, Opt_usrquota
, Opt_grpquota
, Opt_i_version
,
1212 Opt_stripe
, Opt_delalloc
, Opt_nodelalloc
,
1213 Opt_block_validity
, Opt_noblock_validity
,
1214 Opt_inode_readahead_blks
, Opt_journal_ioprio
,
1215 Opt_dioread_nolock
, Opt_dioread_lock
,
1216 Opt_discard
, Opt_nodiscard
,
1219 static const match_table_t tokens
= {
1220 {Opt_bsd_df
, "bsddf"},
1221 {Opt_minix_df
, "minixdf"},
1222 {Opt_grpid
, "grpid"},
1223 {Opt_grpid
, "bsdgroups"},
1224 {Opt_nogrpid
, "nogrpid"},
1225 {Opt_nogrpid
, "sysvgroups"},
1226 {Opt_resgid
, "resgid=%u"},
1227 {Opt_resuid
, "resuid=%u"},
1229 {Opt_err_cont
, "errors=continue"},
1230 {Opt_err_panic
, "errors=panic"},
1231 {Opt_err_ro
, "errors=remount-ro"},
1232 {Opt_nouid32
, "nouid32"},
1233 {Opt_debug
, "debug"},
1234 {Opt_oldalloc
, "oldalloc"},
1235 {Opt_orlov
, "orlov"},
1236 {Opt_user_xattr
, "user_xattr"},
1237 {Opt_nouser_xattr
, "nouser_xattr"},
1239 {Opt_noacl
, "noacl"},
1240 {Opt_noload
, "noload"},
1241 {Opt_noload
, "norecovery"},
1244 {Opt_commit
, "commit=%u"},
1245 {Opt_min_batch_time
, "min_batch_time=%u"},
1246 {Opt_max_batch_time
, "max_batch_time=%u"},
1247 {Opt_journal_update
, "journal=update"},
1248 {Opt_journal_dev
, "journal_dev=%u"},
1249 {Opt_journal_checksum
, "journal_checksum"},
1250 {Opt_journal_async_commit
, "journal_async_commit"},
1251 {Opt_abort
, "abort"},
1252 {Opt_data_journal
, "data=journal"},
1253 {Opt_data_ordered
, "data=ordered"},
1254 {Opt_data_writeback
, "data=writeback"},
1255 {Opt_data_err_abort
, "data_err=abort"},
1256 {Opt_data_err_ignore
, "data_err=ignore"},
1257 {Opt_offusrjquota
, "usrjquota="},
1258 {Opt_usrjquota
, "usrjquota=%s"},
1259 {Opt_offgrpjquota
, "grpjquota="},
1260 {Opt_grpjquota
, "grpjquota=%s"},
1261 {Opt_jqfmt_vfsold
, "jqfmt=vfsold"},
1262 {Opt_jqfmt_vfsv0
, "jqfmt=vfsv0"},
1263 {Opt_jqfmt_vfsv1
, "jqfmt=vfsv1"},
1264 {Opt_grpquota
, "grpquota"},
1265 {Opt_noquota
, "noquota"},
1266 {Opt_quota
, "quota"},
1267 {Opt_usrquota
, "usrquota"},
1268 {Opt_barrier
, "barrier=%u"},
1269 {Opt_barrier
, "barrier"},
1270 {Opt_nobarrier
, "nobarrier"},
1271 {Opt_i_version
, "i_version"},
1272 {Opt_stripe
, "stripe=%u"},
1273 {Opt_resize
, "resize"},
1274 {Opt_delalloc
, "delalloc"},
1275 {Opt_nodelalloc
, "nodelalloc"},
1276 {Opt_block_validity
, "block_validity"},
1277 {Opt_noblock_validity
, "noblock_validity"},
1278 {Opt_inode_readahead_blks
, "inode_readahead_blks=%u"},
1279 {Opt_journal_ioprio
, "journal_ioprio=%u"},
1280 {Opt_auto_da_alloc
, "auto_da_alloc=%u"},
1281 {Opt_auto_da_alloc
, "auto_da_alloc"},
1282 {Opt_noauto_da_alloc
, "noauto_da_alloc"},
1283 {Opt_dioread_nolock
, "dioread_nolock"},
1284 {Opt_dioread_lock
, "dioread_lock"},
1285 {Opt_discard
, "discard"},
1286 {Opt_nodiscard
, "nodiscard"},
1290 static ext4_fsblk_t
get_sb_block(void **data
)
1292 ext4_fsblk_t sb_block
;
1293 char *options
= (char *) *data
;
1295 if (!options
|| strncmp(options
, "sb=", 3) != 0)
1296 return 1; /* Default location */
1299 /* TODO: use simple_strtoll with >32bit ext4 */
1300 sb_block
= simple_strtoul(options
, &options
, 0);
1301 if (*options
&& *options
!= ',') {
1302 printk(KERN_ERR
"EXT4-fs: Invalid sb specification: %s\n",
1306 if (*options
== ',')
1308 *data
= (void *) options
;
1313 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1314 static char deprecated_msg
[] = "Mount option \"%s\" will be removed by %s\n"
1315 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1318 static int set_qf_name(struct super_block
*sb
, int qtype
, substring_t
*args
)
1320 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1323 if (sb_any_quota_loaded(sb
) &&
1324 !sbi
->s_qf_names
[qtype
]) {
1325 ext4_msg(sb
, KERN_ERR
,
1326 "Cannot change journaled "
1327 "quota options when quota turned on");
1330 qname
= match_strdup(args
);
1332 ext4_msg(sb
, KERN_ERR
,
1333 "Not enough memory for storing quotafile name");
1336 if (sbi
->s_qf_names
[qtype
] &&
1337 strcmp(sbi
->s_qf_names
[qtype
], qname
)) {
1338 ext4_msg(sb
, KERN_ERR
,
1339 "%s quota file already specified", QTYPE2NAME(qtype
));
1343 sbi
->s_qf_names
[qtype
] = qname
;
1344 if (strchr(sbi
->s_qf_names
[qtype
], '/')) {
1345 ext4_msg(sb
, KERN_ERR
,
1346 "quotafile must be on filesystem root");
1347 kfree(sbi
->s_qf_names
[qtype
]);
1348 sbi
->s_qf_names
[qtype
] = NULL
;
1351 set_opt(sbi
->s_mount_opt
, QUOTA
);
1355 static int clear_qf_name(struct super_block
*sb
, int qtype
)
1358 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1360 if (sb_any_quota_loaded(sb
) &&
1361 sbi
->s_qf_names
[qtype
]) {
1362 ext4_msg(sb
, KERN_ERR
, "Cannot change journaled quota options"
1363 " when quota turned on");
1367 * The space will be released later when all options are confirmed
1370 sbi
->s_qf_names
[qtype
] = NULL
;
1375 static int parse_options(char *options
, struct super_block
*sb
,
1376 unsigned long *journal_devnum
,
1377 unsigned int *journal_ioprio
,
1378 ext4_fsblk_t
*n_blocks_count
, int is_remount
)
1380 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1382 substring_t args
[MAX_OPT_ARGS
];
1392 while ((p
= strsep(&options
, ",")) != NULL
) {
1398 * Initialize args struct so we know whether arg was
1399 * found; some options take optional arguments.
1401 args
[0].to
= args
[0].from
= 0;
1402 token
= match_token(p
, tokens
, args
);
1405 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1406 clear_opt(sbi
->s_mount_opt
, MINIX_DF
);
1409 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1410 set_opt(sbi
->s_mount_opt
, MINIX_DF
);
1414 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1415 set_opt(sbi
->s_mount_opt
, GRPID
);
1419 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, p
, "2.6.38");
1420 clear_opt(sbi
->s_mount_opt
, GRPID
);
1424 if (match_int(&args
[0], &option
))
1426 sbi
->s_resuid
= option
;
1429 if (match_int(&args
[0], &option
))
1431 sbi
->s_resgid
= option
;
1434 /* handled by get_sb_block() instead of here */
1435 /* *sb_block = match_int(&args[0]); */
1438 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1439 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1440 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1443 clear_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1444 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1445 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1448 clear_opt(sbi
->s_mount_opt
, ERRORS_RO
);
1449 clear_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
1450 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
1453 set_opt(sbi
->s_mount_opt
, NO_UID32
);
1456 set_opt(sbi
->s_mount_opt
, DEBUG
);
1459 set_opt(sbi
->s_mount_opt
, OLDALLOC
);
1462 clear_opt(sbi
->s_mount_opt
, OLDALLOC
);
1464 #ifdef CONFIG_EXT4_FS_XATTR
1465 case Opt_user_xattr
:
1466 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
1468 case Opt_nouser_xattr
:
1469 clear_opt(sbi
->s_mount_opt
, XATTR_USER
);
1472 case Opt_user_xattr
:
1473 case Opt_nouser_xattr
:
1474 ext4_msg(sb
, KERN_ERR
, "(no)user_xattr options not supported");
1477 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1479 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1482 clear_opt(sbi
->s_mount_opt
, POSIX_ACL
);
1487 ext4_msg(sb
, KERN_ERR
, "(no)acl options not supported");
1490 case Opt_journal_update
:
1492 /* Eventually we will want to be able to create
1493 a journal file here. For now, only allow the
1494 user to specify an existing inode to be the
1497 ext4_msg(sb
, KERN_ERR
,
1498 "Cannot specify journal on remount");
1501 set_opt(sbi
->s_mount_opt
, UPDATE_JOURNAL
);
1503 case Opt_journal_dev
:
1505 ext4_msg(sb
, KERN_ERR
,
1506 "Cannot specify journal on remount");
1509 if (match_int(&args
[0], &option
))
1511 *journal_devnum
= option
;
1513 case Opt_journal_checksum
:
1514 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1516 case Opt_journal_async_commit
:
1517 set_opt(sbi
->s_mount_opt
, JOURNAL_ASYNC_COMMIT
);
1518 set_opt(sbi
->s_mount_opt
, JOURNAL_CHECKSUM
);
1521 set_opt(sbi
->s_mount_opt
, NOLOAD
);
1524 if (match_int(&args
[0], &option
))
1529 option
= JBD2_DEFAULT_MAX_COMMIT_AGE
;
1530 sbi
->s_commit_interval
= HZ
* option
;
1532 case Opt_max_batch_time
:
1533 if (match_int(&args
[0], &option
))
1538 option
= EXT4_DEF_MAX_BATCH_TIME
;
1539 sbi
->s_max_batch_time
= option
;
1541 case Opt_min_batch_time
:
1542 if (match_int(&args
[0], &option
))
1546 sbi
->s_min_batch_time
= option
;
1548 case Opt_data_journal
:
1549 data_opt
= EXT4_MOUNT_JOURNAL_DATA
;
1551 case Opt_data_ordered
:
1552 data_opt
= EXT4_MOUNT_ORDERED_DATA
;
1554 case Opt_data_writeback
:
1555 data_opt
= EXT4_MOUNT_WRITEBACK_DATA
;
1558 if (test_opt(sb
, DATA_FLAGS
) != data_opt
) {
1559 ext4_msg(sb
, KERN_ERR
,
1560 "Cannot change data mode on remount");
1564 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
1565 sbi
->s_mount_opt
|= data_opt
;
1568 case Opt_data_err_abort
:
1569 set_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1571 case Opt_data_err_ignore
:
1572 clear_opt(sbi
->s_mount_opt
, DATA_ERR_ABORT
);
1576 if (!set_qf_name(sb
, USRQUOTA
, &args
[0]))
1580 if (!set_qf_name(sb
, GRPQUOTA
, &args
[0]))
1583 case Opt_offusrjquota
:
1584 if (!clear_qf_name(sb
, USRQUOTA
))
1587 case Opt_offgrpjquota
:
1588 if (!clear_qf_name(sb
, GRPQUOTA
))
1592 case Opt_jqfmt_vfsold
:
1593 qfmt
= QFMT_VFS_OLD
;
1595 case Opt_jqfmt_vfsv0
:
1598 case Opt_jqfmt_vfsv1
:
1601 if (sb_any_quota_loaded(sb
) &&
1602 sbi
->s_jquota_fmt
!= qfmt
) {
1603 ext4_msg(sb
, KERN_ERR
, "Cannot change "
1604 "journaled quota options when "
1608 sbi
->s_jquota_fmt
= qfmt
;
1612 set_opt(sbi
->s_mount_opt
, QUOTA
);
1613 set_opt(sbi
->s_mount_opt
, USRQUOTA
);
1616 set_opt(sbi
->s_mount_opt
, QUOTA
);
1617 set_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1620 if (sb_any_quota_loaded(sb
)) {
1621 ext4_msg(sb
, KERN_ERR
, "Cannot change quota "
1622 "options when quota turned on");
1625 clear_opt(sbi
->s_mount_opt
, QUOTA
);
1626 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1627 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1633 ext4_msg(sb
, KERN_ERR
,
1634 "quota options not supported");
1638 case Opt_offusrjquota
:
1639 case Opt_offgrpjquota
:
1640 case Opt_jqfmt_vfsold
:
1641 case Opt_jqfmt_vfsv0
:
1642 case Opt_jqfmt_vfsv1
:
1643 ext4_msg(sb
, KERN_ERR
,
1644 "journaled quota options not supported");
1650 sbi
->s_mount_flags
|= EXT4_MF_FS_ABORTED
;
1653 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1657 if (match_int(&args
[0], &option
))
1660 option
= 1; /* No argument, default to 1 */
1662 set_opt(sbi
->s_mount_opt
, BARRIER
);
1664 clear_opt(sbi
->s_mount_opt
, BARRIER
);
1670 ext4_msg(sb
, KERN_ERR
,
1671 "resize option only available "
1675 if (match_int(&args
[0], &option
) != 0)
1677 *n_blocks_count
= option
;
1680 ext4_msg(sb
, KERN_WARNING
,
1681 "Ignoring deprecated nobh option");
1684 ext4_msg(sb
, KERN_WARNING
,
1685 "Ignoring deprecated bh option");
1688 set_opt(sbi
->s_mount_opt
, I_VERSION
);
1689 sb
->s_flags
|= MS_I_VERSION
;
1691 case Opt_nodelalloc
:
1692 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
1695 if (match_int(&args
[0], &option
))
1699 sbi
->s_stripe
= option
;
1702 set_opt(sbi
->s_mount_opt
, DELALLOC
);
1704 case Opt_block_validity
:
1705 set_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1707 case Opt_noblock_validity
:
1708 clear_opt(sbi
->s_mount_opt
, BLOCK_VALIDITY
);
1710 case Opt_inode_readahead_blks
:
1711 if (match_int(&args
[0], &option
))
1713 if (option
< 0 || option
> (1 << 30))
1715 if (!is_power_of_2(option
)) {
1716 ext4_msg(sb
, KERN_ERR
,
1717 "EXT4-fs: inode_readahead_blks"
1718 " must be a power of 2");
1721 sbi
->s_inode_readahead_blks
= option
;
1723 case Opt_journal_ioprio
:
1724 if (match_int(&args
[0], &option
))
1726 if (option
< 0 || option
> 7)
1728 *journal_ioprio
= IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE
,
1731 case Opt_noauto_da_alloc
:
1732 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1734 case Opt_auto_da_alloc
:
1736 if (match_int(&args
[0], &option
))
1739 option
= 1; /* No argument, default to 1 */
1741 clear_opt(sbi
->s_mount_opt
, NO_AUTO_DA_ALLOC
);
1743 set_opt(sbi
->s_mount_opt
,NO_AUTO_DA_ALLOC
);
1746 set_opt(sbi
->s_mount_opt
, DISCARD
);
1749 clear_opt(sbi
->s_mount_opt
, DISCARD
);
1751 case Opt_dioread_nolock
:
1752 set_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
1754 case Opt_dioread_lock
:
1755 clear_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
1758 ext4_msg(sb
, KERN_ERR
,
1759 "Unrecognized mount option \"%s\" "
1760 "or missing value", p
);
1765 if (sbi
->s_qf_names
[USRQUOTA
] || sbi
->s_qf_names
[GRPQUOTA
]) {
1766 if (test_opt(sb
, USRQUOTA
) && sbi
->s_qf_names
[USRQUOTA
])
1767 clear_opt(sbi
->s_mount_opt
, USRQUOTA
);
1769 if (test_opt(sb
, GRPQUOTA
) && sbi
->s_qf_names
[GRPQUOTA
])
1770 clear_opt(sbi
->s_mount_opt
, GRPQUOTA
);
1772 if (test_opt(sb
, GRPQUOTA
) || test_opt(sb
, USRQUOTA
)) {
1773 ext4_msg(sb
, KERN_ERR
, "old and new quota "
1778 if (!sbi
->s_jquota_fmt
) {
1779 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1784 if (sbi
->s_jquota_fmt
) {
1785 ext4_msg(sb
, KERN_ERR
, "journaled quota format "
1786 "specified with no journaling "
1795 static int ext4_setup_super(struct super_block
*sb
, struct ext4_super_block
*es
,
1798 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1801 if (le32_to_cpu(es
->s_rev_level
) > EXT4_MAX_SUPP_REV
) {
1802 ext4_msg(sb
, KERN_ERR
, "revision level too high, "
1803 "forcing read-only mode");
1808 if (!(sbi
->s_mount_state
& EXT4_VALID_FS
))
1809 ext4_msg(sb
, KERN_WARNING
, "warning: mounting unchecked fs, "
1810 "running e2fsck is recommended");
1811 else if ((sbi
->s_mount_state
& EXT4_ERROR_FS
))
1812 ext4_msg(sb
, KERN_WARNING
,
1813 "warning: mounting fs with errors, "
1814 "running e2fsck is recommended");
1815 else if ((__s16
) le16_to_cpu(es
->s_max_mnt_count
) >= 0 &&
1816 le16_to_cpu(es
->s_mnt_count
) >=
1817 (unsigned short) (__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1818 ext4_msg(sb
, KERN_WARNING
,
1819 "warning: maximal mount count reached, "
1820 "running e2fsck is recommended");
1821 else if (le32_to_cpu(es
->s_checkinterval
) &&
1822 (le32_to_cpu(es
->s_lastcheck
) +
1823 le32_to_cpu(es
->s_checkinterval
) <= get_seconds()))
1824 ext4_msg(sb
, KERN_WARNING
,
1825 "warning: checktime reached, "
1826 "running e2fsck is recommended");
1827 if (!sbi
->s_journal
)
1828 es
->s_state
&= cpu_to_le16(~EXT4_VALID_FS
);
1829 if (!(__s16
) le16_to_cpu(es
->s_max_mnt_count
))
1830 es
->s_max_mnt_count
= cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT
);
1831 le16_add_cpu(&es
->s_mnt_count
, 1);
1832 es
->s_mtime
= cpu_to_le32(get_seconds());
1833 ext4_update_dynamic_rev(sb
);
1835 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
1837 ext4_commit_super(sb
, 1);
1838 if (test_opt(sb
, DEBUG
))
1839 printk(KERN_INFO
"[EXT4 FS bs=%lu, gc=%u, "
1840 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1842 sbi
->s_groups_count
,
1843 EXT4_BLOCKS_PER_GROUP(sb
),
1844 EXT4_INODES_PER_GROUP(sb
),
1850 static int ext4_fill_flex_info(struct super_block
*sb
)
1852 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1853 struct ext4_group_desc
*gdp
= NULL
;
1854 ext4_group_t flex_group_count
;
1855 ext4_group_t flex_group
;
1856 int groups_per_flex
= 0;
1860 sbi
->s_log_groups_per_flex
= sbi
->s_es
->s_log_groups_per_flex
;
1861 groups_per_flex
= 1 << sbi
->s_log_groups_per_flex
;
1863 if (groups_per_flex
< 2) {
1864 sbi
->s_log_groups_per_flex
= 0;
1868 /* We allocate both existing and potentially added groups */
1869 flex_group_count
= ((sbi
->s_groups_count
+ groups_per_flex
- 1) +
1870 ((le16_to_cpu(sbi
->s_es
->s_reserved_gdt_blocks
) + 1) <<
1871 EXT4_DESC_PER_BLOCK_BITS(sb
))) / groups_per_flex
;
1872 size
= flex_group_count
* sizeof(struct flex_groups
);
1873 sbi
->s_flex_groups
= kzalloc(size
, GFP_KERNEL
);
1874 if (sbi
->s_flex_groups
== NULL
) {
1875 sbi
->s_flex_groups
= vmalloc(size
);
1876 if (sbi
->s_flex_groups
)
1877 memset(sbi
->s_flex_groups
, 0, size
);
1879 if (sbi
->s_flex_groups
== NULL
) {
1880 ext4_msg(sb
, KERN_ERR
, "not enough memory for "
1881 "%u flex groups", flex_group_count
);
1885 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1886 gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1888 flex_group
= ext4_flex_group(sbi
, i
);
1889 atomic_add(ext4_free_inodes_count(sb
, gdp
),
1890 &sbi
->s_flex_groups
[flex_group
].free_inodes
);
1891 atomic_add(ext4_free_blks_count(sb
, gdp
),
1892 &sbi
->s_flex_groups
[flex_group
].free_blocks
);
1893 atomic_add(ext4_used_dirs_count(sb
, gdp
),
1894 &sbi
->s_flex_groups
[flex_group
].used_dirs
);
1902 __le16
ext4_group_desc_csum(struct ext4_sb_info
*sbi
, __u32 block_group
,
1903 struct ext4_group_desc
*gdp
)
1907 if (sbi
->s_es
->s_feature_ro_compat
&
1908 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) {
1909 int offset
= offsetof(struct ext4_group_desc
, bg_checksum
);
1910 __le32 le_group
= cpu_to_le32(block_group
);
1912 crc
= crc16(~0, sbi
->s_es
->s_uuid
, sizeof(sbi
->s_es
->s_uuid
));
1913 crc
= crc16(crc
, (__u8
*)&le_group
, sizeof(le_group
));
1914 crc
= crc16(crc
, (__u8
*)gdp
, offset
);
1915 offset
+= sizeof(gdp
->bg_checksum
); /* skip checksum */
1916 /* for checksum of struct ext4_group_desc do the rest...*/
1917 if ((sbi
->s_es
->s_feature_incompat
&
1918 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT
)) &&
1919 offset
< le16_to_cpu(sbi
->s_es
->s_desc_size
))
1920 crc
= crc16(crc
, (__u8
*)gdp
+ offset
,
1921 le16_to_cpu(sbi
->s_es
->s_desc_size
) -
1925 return cpu_to_le16(crc
);
1928 int ext4_group_desc_csum_verify(struct ext4_sb_info
*sbi
, __u32 block_group
,
1929 struct ext4_group_desc
*gdp
)
1931 if ((sbi
->s_es
->s_feature_ro_compat
&
1932 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM
)) &&
1933 (gdp
->bg_checksum
!= ext4_group_desc_csum(sbi
, block_group
, gdp
)))
1939 /* Called at mount-time, super-block is locked */
1940 static int ext4_check_descriptors(struct super_block
*sb
)
1942 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
1943 ext4_fsblk_t first_block
= le32_to_cpu(sbi
->s_es
->s_first_data_block
);
1944 ext4_fsblk_t last_block
;
1945 ext4_fsblk_t block_bitmap
;
1946 ext4_fsblk_t inode_bitmap
;
1947 ext4_fsblk_t inode_table
;
1948 int flexbg_flag
= 0;
1951 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
1954 ext4_debug("Checking group descriptors");
1956 for (i
= 0; i
< sbi
->s_groups_count
; i
++) {
1957 struct ext4_group_desc
*gdp
= ext4_get_group_desc(sb
, i
, NULL
);
1959 if (i
== sbi
->s_groups_count
- 1 || flexbg_flag
)
1960 last_block
= ext4_blocks_count(sbi
->s_es
) - 1;
1962 last_block
= first_block
+
1963 (EXT4_BLOCKS_PER_GROUP(sb
) - 1);
1965 block_bitmap
= ext4_block_bitmap(sb
, gdp
);
1966 if (block_bitmap
< first_block
|| block_bitmap
> last_block
) {
1967 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1968 "Block bitmap for group %u not in group "
1969 "(block %llu)!", i
, block_bitmap
);
1972 inode_bitmap
= ext4_inode_bitmap(sb
, gdp
);
1973 if (inode_bitmap
< first_block
|| inode_bitmap
> last_block
) {
1974 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1975 "Inode bitmap for group %u not in group "
1976 "(block %llu)!", i
, inode_bitmap
);
1979 inode_table
= ext4_inode_table(sb
, gdp
);
1980 if (inode_table
< first_block
||
1981 inode_table
+ sbi
->s_itb_per_group
- 1 > last_block
) {
1982 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1983 "Inode table for group %u not in group "
1984 "(block %llu)!", i
, inode_table
);
1987 ext4_lock_group(sb
, i
);
1988 if (!ext4_group_desc_csum_verify(sbi
, i
, gdp
)) {
1989 ext4_msg(sb
, KERN_ERR
, "ext4_check_descriptors: "
1990 "Checksum for group %u failed (%u!=%u)",
1991 i
, le16_to_cpu(ext4_group_desc_csum(sbi
, i
,
1992 gdp
)), le16_to_cpu(gdp
->bg_checksum
));
1993 if (!(sb
->s_flags
& MS_RDONLY
)) {
1994 ext4_unlock_group(sb
, i
);
1998 ext4_unlock_group(sb
, i
);
2000 first_block
+= EXT4_BLOCKS_PER_GROUP(sb
);
2003 ext4_free_blocks_count_set(sbi
->s_es
, ext4_count_free_blocks(sb
));
2004 sbi
->s_es
->s_free_inodes_count
=cpu_to_le32(ext4_count_free_inodes(sb
));
2008 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2009 * the superblock) which were deleted from all directories, but held open by
2010 * a process at the time of a crash. We walk the list and try to delete these
2011 * inodes at recovery time (only with a read-write filesystem).
2013 * In order to keep the orphan inode chain consistent during traversal (in
2014 * case of crash during recovery), we link each inode into the superblock
2015 * orphan list_head and handle it the same way as an inode deletion during
2016 * normal operation (which journals the operations for us).
2018 * We only do an iget() and an iput() on each inode, which is very safe if we
2019 * accidentally point at an in-use or already deleted inode. The worst that
2020 * can happen in this case is that we get a "bit already cleared" message from
2021 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2022 * e2fsck was run on this filesystem, and it must have already done the orphan
2023 * inode cleanup for us, so we can safely abort without any further action.
2025 static void ext4_orphan_cleanup(struct super_block
*sb
,
2026 struct ext4_super_block
*es
)
2028 unsigned int s_flags
= sb
->s_flags
;
2029 int nr_orphans
= 0, nr_truncates
= 0;
2033 if (!es
->s_last_orphan
) {
2034 jbd_debug(4, "no orphan inodes to clean up\n");
2038 if (bdev_read_only(sb
->s_bdev
)) {
2039 ext4_msg(sb
, KERN_ERR
, "write access "
2040 "unavailable, skipping orphan cleanup");
2044 if (EXT4_SB(sb
)->s_mount_state
& EXT4_ERROR_FS
) {
2045 if (es
->s_last_orphan
)
2046 jbd_debug(1, "Errors on filesystem, "
2047 "clearing orphan list.\n");
2048 es
->s_last_orphan
= 0;
2049 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2053 if (s_flags
& MS_RDONLY
) {
2054 ext4_msg(sb
, KERN_INFO
, "orphan cleanup on readonly fs");
2055 sb
->s_flags
&= ~MS_RDONLY
;
2058 /* Needed for iput() to work correctly and not trash data */
2059 sb
->s_flags
|= MS_ACTIVE
;
2060 /* Turn on quotas so that they are updated correctly */
2061 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2062 if (EXT4_SB(sb
)->s_qf_names
[i
]) {
2063 int ret
= ext4_quota_on_mount(sb
, i
);
2065 ext4_msg(sb
, KERN_ERR
,
2066 "Cannot turn on journaled "
2067 "quota: error %d", ret
);
2072 while (es
->s_last_orphan
) {
2073 struct inode
*inode
;
2075 inode
= ext4_orphan_get(sb
, le32_to_cpu(es
->s_last_orphan
));
2076 if (IS_ERR(inode
)) {
2077 es
->s_last_orphan
= 0;
2081 list_add(&EXT4_I(inode
)->i_orphan
, &EXT4_SB(sb
)->s_orphan
);
2082 dquot_initialize(inode
);
2083 if (inode
->i_nlink
) {
2084 ext4_msg(sb
, KERN_DEBUG
,
2085 "%s: truncating inode %lu to %lld bytes",
2086 __func__
, inode
->i_ino
, inode
->i_size
);
2087 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2088 inode
->i_ino
, inode
->i_size
);
2089 ext4_truncate(inode
);
2092 ext4_msg(sb
, KERN_DEBUG
,
2093 "%s: deleting unreferenced inode %lu",
2094 __func__
, inode
->i_ino
);
2095 jbd_debug(2, "deleting unreferenced inode %lu\n",
2099 iput(inode
); /* The delete magic happens here! */
2102 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2105 ext4_msg(sb
, KERN_INFO
, "%d orphan inode%s deleted",
2106 PLURAL(nr_orphans
));
2108 ext4_msg(sb
, KERN_INFO
, "%d truncate%s cleaned up",
2109 PLURAL(nr_truncates
));
2111 /* Turn quotas off */
2112 for (i
= 0; i
< MAXQUOTAS
; i
++) {
2113 if (sb_dqopt(sb
)->files
[i
])
2114 dquot_quota_off(sb
, i
);
2117 sb
->s_flags
= s_flags
; /* Restore MS_RDONLY status */
2121 * Maximal extent format file size.
2122 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2123 * extent format containers, within a sector_t, and within i_blocks
2124 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2125 * so that won't be a limiting factor.
2127 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2129 static loff_t
ext4_max_size(int blkbits
, int has_huge_files
)
2132 loff_t upper_limit
= MAX_LFS_FILESIZE
;
2134 /* small i_blocks in vfs inode? */
2135 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2137 * CONFIG_LBDAF is not enabled implies the inode
2138 * i_block represent total blocks in 512 bytes
2139 * 32 == size of vfs inode i_blocks * 8
2141 upper_limit
= (1LL << 32) - 1;
2143 /* total blocks in file system block size */
2144 upper_limit
>>= (blkbits
- 9);
2145 upper_limit
<<= blkbits
;
2148 /* 32-bit extent-start container, ee_block */
2153 /* Sanity check against vm- & vfs- imposed limits */
2154 if (res
> upper_limit
)
2161 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2162 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2163 * We need to be 1 filesystem block less than the 2^48 sector limit.
2165 static loff_t
ext4_max_bitmap_size(int bits
, int has_huge_files
)
2167 loff_t res
= EXT4_NDIR_BLOCKS
;
2170 /* This is calculated to be the largest file size for a dense, block
2171 * mapped file such that the file's total number of 512-byte sectors,
2172 * including data and all indirect blocks, does not exceed (2^48 - 1).
2174 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2175 * number of 512-byte sectors of the file.
2178 if (!has_huge_files
|| sizeof(blkcnt_t
) < sizeof(u64
)) {
2180 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2181 * the inode i_block field represents total file blocks in
2182 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2184 upper_limit
= (1LL << 32) - 1;
2186 /* total blocks in file system block size */
2187 upper_limit
>>= (bits
- 9);
2191 * We use 48 bit ext4_inode i_blocks
2192 * With EXT4_HUGE_FILE_FL set the i_blocks
2193 * represent total number of blocks in
2194 * file system block size
2196 upper_limit
= (1LL << 48) - 1;
2200 /* indirect blocks */
2202 /* double indirect blocks */
2203 meta_blocks
+= 1 + (1LL << (bits
-2));
2204 /* tripple indirect blocks */
2205 meta_blocks
+= 1 + (1LL << (bits
-2)) + (1LL << (2*(bits
-2)));
2207 upper_limit
-= meta_blocks
;
2208 upper_limit
<<= bits
;
2210 res
+= 1LL << (bits
-2);
2211 res
+= 1LL << (2*(bits
-2));
2212 res
+= 1LL << (3*(bits
-2));
2214 if (res
> upper_limit
)
2217 if (res
> MAX_LFS_FILESIZE
)
2218 res
= MAX_LFS_FILESIZE
;
2223 static ext4_fsblk_t
descriptor_loc(struct super_block
*sb
,
2224 ext4_fsblk_t logical_sb_block
, int nr
)
2226 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
2227 ext4_group_t bg
, first_meta_bg
;
2230 first_meta_bg
= le32_to_cpu(sbi
->s_es
->s_first_meta_bg
);
2232 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_META_BG
) ||
2234 return logical_sb_block
+ nr
+ 1;
2235 bg
= sbi
->s_desc_per_block
* nr
;
2236 if (ext4_bg_has_super(sb
, bg
))
2239 return (has_super
+ ext4_group_first_block_no(sb
, bg
));
2243 * ext4_get_stripe_size: Get the stripe size.
2244 * @sbi: In memory super block info
2246 * If we have specified it via mount option, then
2247 * use the mount option value. If the value specified at mount time is
2248 * greater than the blocks per group use the super block value.
2249 * If the super block value is greater than blocks per group return 0.
2250 * Allocator needs it be less than blocks per group.
2253 static unsigned long ext4_get_stripe_size(struct ext4_sb_info
*sbi
)
2255 unsigned long stride
= le16_to_cpu(sbi
->s_es
->s_raid_stride
);
2256 unsigned long stripe_width
=
2257 le32_to_cpu(sbi
->s_es
->s_raid_stripe_width
);
2259 if (sbi
->s_stripe
&& sbi
->s_stripe
<= sbi
->s_blocks_per_group
)
2260 return sbi
->s_stripe
;
2262 if (stripe_width
<= sbi
->s_blocks_per_group
)
2263 return stripe_width
;
2265 if (stride
<= sbi
->s_blocks_per_group
)
2274 struct attribute attr
;
2275 ssize_t (*show
)(struct ext4_attr
*, struct ext4_sb_info
*, char *);
2276 ssize_t (*store
)(struct ext4_attr
*, struct ext4_sb_info
*,
2277 const char *, size_t);
2281 static int parse_strtoul(const char *buf
,
2282 unsigned long max
, unsigned long *value
)
2286 *value
= simple_strtoul(skip_spaces(buf
), &endp
, 0);
2287 endp
= skip_spaces(endp
);
2288 if (*endp
|| *value
> max
)
2294 static ssize_t
delayed_allocation_blocks_show(struct ext4_attr
*a
,
2295 struct ext4_sb_info
*sbi
,
2298 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2299 (s64
) percpu_counter_sum(&sbi
->s_dirtyblocks_counter
));
2302 static ssize_t
session_write_kbytes_show(struct ext4_attr
*a
,
2303 struct ext4_sb_info
*sbi
, char *buf
)
2305 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2307 return snprintf(buf
, PAGE_SIZE
, "%lu\n",
2308 (part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2309 sbi
->s_sectors_written_start
) >> 1);
2312 static ssize_t
lifetime_write_kbytes_show(struct ext4_attr
*a
,
2313 struct ext4_sb_info
*sbi
, char *buf
)
2315 struct super_block
*sb
= sbi
->s_buddy_cache
->i_sb
;
2317 return snprintf(buf
, PAGE_SIZE
, "%llu\n",
2318 (unsigned long long)(sbi
->s_kbytes_written
+
2319 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
2320 EXT4_SB(sb
)->s_sectors_written_start
) >> 1)));
2323 static ssize_t
inode_readahead_blks_store(struct ext4_attr
*a
,
2324 struct ext4_sb_info
*sbi
,
2325 const char *buf
, size_t count
)
2329 if (parse_strtoul(buf
, 0x40000000, &t
))
2332 if (!is_power_of_2(t
))
2335 sbi
->s_inode_readahead_blks
= t
;
2339 static ssize_t
sbi_ui_show(struct ext4_attr
*a
,
2340 struct ext4_sb_info
*sbi
, char *buf
)
2342 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2344 return snprintf(buf
, PAGE_SIZE
, "%u\n", *ui
);
2347 static ssize_t
sbi_ui_store(struct ext4_attr
*a
,
2348 struct ext4_sb_info
*sbi
,
2349 const char *buf
, size_t count
)
2351 unsigned int *ui
= (unsigned int *) (((char *) sbi
) + a
->offset
);
2354 if (parse_strtoul(buf
, 0xffffffff, &t
))
2360 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2361 static struct ext4_attr ext4_attr_##_name = { \
2362 .attr = {.name = __stringify(_name), .mode = _mode }, \
2365 .offset = offsetof(struct ext4_sb_info, _elname), \
2367 #define EXT4_ATTR(name, mode, show, store) \
2368 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2370 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2371 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2372 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2373 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2374 #define ATTR_LIST(name) &ext4_attr_##name.attr
2376 EXT4_RO_ATTR(delayed_allocation_blocks
);
2377 EXT4_RO_ATTR(session_write_kbytes
);
2378 EXT4_RO_ATTR(lifetime_write_kbytes
);
2379 EXT4_ATTR_OFFSET(inode_readahead_blks
, 0644, sbi_ui_show
,
2380 inode_readahead_blks_store
, s_inode_readahead_blks
);
2381 EXT4_RW_ATTR_SBI_UI(inode_goal
, s_inode_goal
);
2382 EXT4_RW_ATTR_SBI_UI(mb_stats
, s_mb_stats
);
2383 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan
, s_mb_max_to_scan
);
2384 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan
, s_mb_min_to_scan
);
2385 EXT4_RW_ATTR_SBI_UI(mb_order2_req
, s_mb_order2_reqs
);
2386 EXT4_RW_ATTR_SBI_UI(mb_stream_req
, s_mb_stream_request
);
2387 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc
, s_mb_group_prealloc
);
2388 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump
, s_max_writeback_mb_bump
);
2390 static struct attribute
*ext4_attrs
[] = {
2391 ATTR_LIST(delayed_allocation_blocks
),
2392 ATTR_LIST(session_write_kbytes
),
2393 ATTR_LIST(lifetime_write_kbytes
),
2394 ATTR_LIST(inode_readahead_blks
),
2395 ATTR_LIST(inode_goal
),
2396 ATTR_LIST(mb_stats
),
2397 ATTR_LIST(mb_max_to_scan
),
2398 ATTR_LIST(mb_min_to_scan
),
2399 ATTR_LIST(mb_order2_req
),
2400 ATTR_LIST(mb_stream_req
),
2401 ATTR_LIST(mb_group_prealloc
),
2402 ATTR_LIST(max_writeback_mb_bump
),
2406 static ssize_t
ext4_attr_show(struct kobject
*kobj
,
2407 struct attribute
*attr
, char *buf
)
2409 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2411 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2413 return a
->show
? a
->show(a
, sbi
, buf
) : 0;
2416 static ssize_t
ext4_attr_store(struct kobject
*kobj
,
2417 struct attribute
*attr
,
2418 const char *buf
, size_t len
)
2420 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2422 struct ext4_attr
*a
= container_of(attr
, struct ext4_attr
, attr
);
2424 return a
->store
? a
->store(a
, sbi
, buf
, len
) : 0;
2427 static void ext4_sb_release(struct kobject
*kobj
)
2429 struct ext4_sb_info
*sbi
= container_of(kobj
, struct ext4_sb_info
,
2431 complete(&sbi
->s_kobj_unregister
);
2435 static const struct sysfs_ops ext4_attr_ops
= {
2436 .show
= ext4_attr_show
,
2437 .store
= ext4_attr_store
,
2440 static struct kobj_type ext4_ktype
= {
2441 .default_attrs
= ext4_attrs
,
2442 .sysfs_ops
= &ext4_attr_ops
,
2443 .release
= ext4_sb_release
,
2447 * Check whether this filesystem can be mounted based on
2448 * the features present and the RDONLY/RDWR mount requested.
2449 * Returns 1 if this filesystem can be mounted as requested,
2450 * 0 if it cannot be.
2452 static int ext4_feature_set_ok(struct super_block
*sb
, int readonly
)
2454 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, ~EXT4_FEATURE_INCOMPAT_SUPP
)) {
2455 ext4_msg(sb
, KERN_ERR
,
2456 "Couldn't mount because of "
2457 "unsupported optional features (%x)",
2458 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_incompat
) &
2459 ~EXT4_FEATURE_INCOMPAT_SUPP
));
2466 /* Check that feature set is OK for a read-write mount */
2467 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~EXT4_FEATURE_RO_COMPAT_SUPP
)) {
2468 ext4_msg(sb
, KERN_ERR
, "couldn't mount RDWR because of "
2469 "unsupported optional features (%x)",
2470 (le32_to_cpu(EXT4_SB(sb
)->s_es
->s_feature_ro_compat
) &
2471 ~EXT4_FEATURE_RO_COMPAT_SUPP
));
2475 * Large file size enabled file system can only be mounted
2476 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2478 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
, EXT4_FEATURE_RO_COMPAT_HUGE_FILE
)) {
2479 if (sizeof(blkcnt_t
) < sizeof(u64
)) {
2480 ext4_msg(sb
, KERN_ERR
, "Filesystem with huge files "
2481 "cannot be mounted RDWR without "
2490 * This function is called once a day if we have errors logged
2491 * on the file system
2493 static void print_daily_error_info(unsigned long arg
)
2495 struct super_block
*sb
= (struct super_block
*) arg
;
2496 struct ext4_sb_info
*sbi
;
2497 struct ext4_super_block
*es
;
2502 if (es
->s_error_count
)
2503 ext4_msg(sb
, KERN_NOTICE
, "error count: %u",
2504 le32_to_cpu(es
->s_error_count
));
2505 if (es
->s_first_error_time
) {
2506 printk(KERN_NOTICE
"EXT4-fs (%s): initial error at %u: %.*s:%d",
2507 sb
->s_id
, le32_to_cpu(es
->s_first_error_time
),
2508 (int) sizeof(es
->s_first_error_func
),
2509 es
->s_first_error_func
,
2510 le32_to_cpu(es
->s_first_error_line
));
2511 if (es
->s_first_error_ino
)
2512 printk(": inode %u",
2513 le32_to_cpu(es
->s_first_error_ino
));
2514 if (es
->s_first_error_block
)
2515 printk(": block %llu", (unsigned long long)
2516 le64_to_cpu(es
->s_first_error_block
));
2519 if (es
->s_last_error_time
) {
2520 printk(KERN_NOTICE
"EXT4-fs (%s): last error at %u: %.*s:%d",
2521 sb
->s_id
, le32_to_cpu(es
->s_last_error_time
),
2522 (int) sizeof(es
->s_last_error_func
),
2523 es
->s_last_error_func
,
2524 le32_to_cpu(es
->s_last_error_line
));
2525 if (es
->s_last_error_ino
)
2526 printk(": inode %u",
2527 le32_to_cpu(es
->s_last_error_ino
));
2528 if (es
->s_last_error_block
)
2529 printk(": block %llu", (unsigned long long)
2530 le64_to_cpu(es
->s_last_error_block
));
2533 mod_timer(&sbi
->s_err_report
, jiffies
+ 24*60*60*HZ
); /* Once a day */
2536 static int ext4_fill_super(struct super_block
*sb
, void *data
, int silent
)
2537 __releases(kernel_lock
)
2538 __acquires(kernel_lock
)
2540 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
2541 struct buffer_head
*bh
;
2542 struct ext4_super_block
*es
= NULL
;
2543 struct ext4_sb_info
*sbi
;
2545 ext4_fsblk_t sb_block
= get_sb_block(&data
);
2546 ext4_fsblk_t logical_sb_block
;
2547 unsigned long offset
= 0;
2548 unsigned long journal_devnum
= 0;
2549 unsigned long def_mount_opts
;
2555 unsigned int db_count
;
2557 int needs_recovery
, has_huge_files
;
2560 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
2562 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
2566 sbi
->s_blockgroup_lock
=
2567 kzalloc(sizeof(struct blockgroup_lock
), GFP_KERNEL
);
2568 if (!sbi
->s_blockgroup_lock
) {
2572 sb
->s_fs_info
= sbi
;
2573 sbi
->s_mount_opt
= 0;
2574 sbi
->s_resuid
= EXT4_DEF_RESUID
;
2575 sbi
->s_resgid
= EXT4_DEF_RESGID
;
2576 sbi
->s_inode_readahead_blks
= EXT4_DEF_INODE_READAHEAD_BLKS
;
2577 sbi
->s_sb_block
= sb_block
;
2578 sbi
->s_sectors_written_start
= part_stat_read(sb
->s_bdev
->bd_part
,
2583 /* Cleanup superblock name */
2584 for (cp
= sb
->s_id
; (cp
= strchr(cp
, '/'));)
2588 blocksize
= sb_min_blocksize(sb
, EXT4_MIN_BLOCK_SIZE
);
2590 ext4_msg(sb
, KERN_ERR
, "unable to set blocksize");
2595 * The ext4 superblock will not be buffer aligned for other than 1kB
2596 * block sizes. We need to calculate the offset from buffer start.
2598 if (blocksize
!= EXT4_MIN_BLOCK_SIZE
) {
2599 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2600 offset
= do_div(logical_sb_block
, blocksize
);
2602 logical_sb_block
= sb_block
;
2605 if (!(bh
= sb_bread(sb
, logical_sb_block
))) {
2606 ext4_msg(sb
, KERN_ERR
, "unable to read superblock");
2610 * Note: s_es must be initialized as soon as possible because
2611 * some ext4 macro-instructions depend on its value
2613 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
2615 sb
->s_magic
= le16_to_cpu(es
->s_magic
);
2616 if (sb
->s_magic
!= EXT4_SUPER_MAGIC
)
2618 sbi
->s_kbytes_written
= le64_to_cpu(es
->s_kbytes_written
);
2620 /* Set defaults before we parse the mount options */
2621 def_mount_opts
= le32_to_cpu(es
->s_default_mount_opts
);
2622 if (def_mount_opts
& EXT4_DEFM_DEBUG
)
2623 set_opt(sbi
->s_mount_opt
, DEBUG
);
2624 if (def_mount_opts
& EXT4_DEFM_BSDGROUPS
) {
2625 ext4_msg(sb
, KERN_WARNING
, deprecated_msg
, "bsdgroups",
2627 set_opt(sbi
->s_mount_opt
, GRPID
);
2629 if (def_mount_opts
& EXT4_DEFM_UID16
)
2630 set_opt(sbi
->s_mount_opt
, NO_UID32
);
2631 #ifdef CONFIG_EXT4_FS_XATTR
2632 if (def_mount_opts
& EXT4_DEFM_XATTR_USER
)
2633 set_opt(sbi
->s_mount_opt
, XATTR_USER
);
2635 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2636 if (def_mount_opts
& EXT4_DEFM_ACL
)
2637 set_opt(sbi
->s_mount_opt
, POSIX_ACL
);
2639 if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_DATA
)
2640 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2641 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_ORDERED
)
2642 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2643 else if ((def_mount_opts
& EXT4_DEFM_JMODE
) == EXT4_DEFM_JMODE_WBACK
)
2644 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
2646 if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_PANIC
)
2647 set_opt(sbi
->s_mount_opt
, ERRORS_PANIC
);
2648 else if (le16_to_cpu(sbi
->s_es
->s_errors
) == EXT4_ERRORS_CONTINUE
)
2649 set_opt(sbi
->s_mount_opt
, ERRORS_CONT
);
2651 set_opt(sbi
->s_mount_opt
, ERRORS_RO
);
2653 sbi
->s_resuid
= le16_to_cpu(es
->s_def_resuid
);
2654 sbi
->s_resgid
= le16_to_cpu(es
->s_def_resgid
);
2655 sbi
->s_commit_interval
= JBD2_DEFAULT_MAX_COMMIT_AGE
* HZ
;
2656 sbi
->s_min_batch_time
= EXT4_DEF_MIN_BATCH_TIME
;
2657 sbi
->s_max_batch_time
= EXT4_DEF_MAX_BATCH_TIME
;
2659 set_opt(sbi
->s_mount_opt
, BARRIER
);
2662 * enable delayed allocation by default
2663 * Use -o nodelalloc to turn it off
2665 if (!IS_EXT3_SB(sb
))
2666 set_opt(sbi
->s_mount_opt
, DELALLOC
);
2668 if (!parse_options((char *) data
, sb
, &journal_devnum
,
2669 &journal_ioprio
, NULL
, 0))
2672 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
2673 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
2675 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
&&
2676 (EXT4_HAS_COMPAT_FEATURE(sb
, ~0U) ||
2677 EXT4_HAS_RO_COMPAT_FEATURE(sb
, ~0U) ||
2678 EXT4_HAS_INCOMPAT_FEATURE(sb
, ~0U)))
2679 ext4_msg(sb
, KERN_WARNING
,
2680 "feature flags set on rev 0 fs, "
2681 "running e2fsck is recommended");
2684 * Check feature flags regardless of the revision level, since we
2685 * previously didn't change the revision level when setting the flags,
2686 * so there is a chance incompat flags are set on a rev 0 filesystem.
2688 if (!ext4_feature_set_ok(sb
, (sb
->s_flags
& MS_RDONLY
)))
2691 blocksize
= BLOCK_SIZE
<< le32_to_cpu(es
->s_log_block_size
);
2693 if (blocksize
< EXT4_MIN_BLOCK_SIZE
||
2694 blocksize
> EXT4_MAX_BLOCK_SIZE
) {
2695 ext4_msg(sb
, KERN_ERR
,
2696 "Unsupported filesystem blocksize %d", blocksize
);
2700 if (sb
->s_blocksize
!= blocksize
) {
2701 /* Validate the filesystem blocksize */
2702 if (!sb_set_blocksize(sb
, blocksize
)) {
2703 ext4_msg(sb
, KERN_ERR
, "bad block size %d",
2709 logical_sb_block
= sb_block
* EXT4_MIN_BLOCK_SIZE
;
2710 offset
= do_div(logical_sb_block
, blocksize
);
2711 bh
= sb_bread(sb
, logical_sb_block
);
2713 ext4_msg(sb
, KERN_ERR
,
2714 "Can't read superblock on 2nd try");
2717 es
= (struct ext4_super_block
*)(((char *)bh
->b_data
) + offset
);
2719 if (es
->s_magic
!= cpu_to_le16(EXT4_SUPER_MAGIC
)) {
2720 ext4_msg(sb
, KERN_ERR
,
2721 "Magic mismatch, very weird!");
2726 has_huge_files
= EXT4_HAS_RO_COMPAT_FEATURE(sb
,
2727 EXT4_FEATURE_RO_COMPAT_HUGE_FILE
);
2728 sbi
->s_bitmap_maxbytes
= ext4_max_bitmap_size(sb
->s_blocksize_bits
,
2730 sb
->s_maxbytes
= ext4_max_size(sb
->s_blocksize_bits
, has_huge_files
);
2732 if (le32_to_cpu(es
->s_rev_level
) == EXT4_GOOD_OLD_REV
) {
2733 sbi
->s_inode_size
= EXT4_GOOD_OLD_INODE_SIZE
;
2734 sbi
->s_first_ino
= EXT4_GOOD_OLD_FIRST_INO
;
2736 sbi
->s_inode_size
= le16_to_cpu(es
->s_inode_size
);
2737 sbi
->s_first_ino
= le32_to_cpu(es
->s_first_ino
);
2738 if ((sbi
->s_inode_size
< EXT4_GOOD_OLD_INODE_SIZE
) ||
2739 (!is_power_of_2(sbi
->s_inode_size
)) ||
2740 (sbi
->s_inode_size
> blocksize
)) {
2741 ext4_msg(sb
, KERN_ERR
,
2742 "unsupported inode size: %d",
2746 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
)
2747 sb
->s_time_gran
= 1 << (EXT4_EPOCH_BITS
- 2);
2750 sbi
->s_desc_size
= le16_to_cpu(es
->s_desc_size
);
2751 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_64BIT
)) {
2752 if (sbi
->s_desc_size
< EXT4_MIN_DESC_SIZE_64BIT
||
2753 sbi
->s_desc_size
> EXT4_MAX_DESC_SIZE
||
2754 !is_power_of_2(sbi
->s_desc_size
)) {
2755 ext4_msg(sb
, KERN_ERR
,
2756 "unsupported descriptor size %lu",
2761 sbi
->s_desc_size
= EXT4_MIN_DESC_SIZE
;
2763 sbi
->s_blocks_per_group
= le32_to_cpu(es
->s_blocks_per_group
);
2764 sbi
->s_inodes_per_group
= le32_to_cpu(es
->s_inodes_per_group
);
2765 if (EXT4_INODE_SIZE(sb
) == 0 || EXT4_INODES_PER_GROUP(sb
) == 0)
2768 sbi
->s_inodes_per_block
= blocksize
/ EXT4_INODE_SIZE(sb
);
2769 if (sbi
->s_inodes_per_block
== 0)
2771 sbi
->s_itb_per_group
= sbi
->s_inodes_per_group
/
2772 sbi
->s_inodes_per_block
;
2773 sbi
->s_desc_per_block
= blocksize
/ EXT4_DESC_SIZE(sb
);
2775 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
2776 sbi
->s_addr_per_block_bits
= ilog2(EXT4_ADDR_PER_BLOCK(sb
));
2777 sbi
->s_desc_per_block_bits
= ilog2(EXT4_DESC_PER_BLOCK(sb
));
2779 for (i
= 0; i
< 4; i
++)
2780 sbi
->s_hash_seed
[i
] = le32_to_cpu(es
->s_hash_seed
[i
]);
2781 sbi
->s_def_hash_version
= es
->s_def_hash_version
;
2782 i
= le32_to_cpu(es
->s_flags
);
2783 if (i
& EXT2_FLAGS_UNSIGNED_HASH
)
2784 sbi
->s_hash_unsigned
= 3;
2785 else if ((i
& EXT2_FLAGS_SIGNED_HASH
) == 0) {
2786 #ifdef __CHAR_UNSIGNED__
2787 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH
);
2788 sbi
->s_hash_unsigned
= 3;
2790 es
->s_flags
|= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH
);
2795 if (sbi
->s_blocks_per_group
> blocksize
* 8) {
2796 ext4_msg(sb
, KERN_ERR
,
2797 "#blocks per group too big: %lu",
2798 sbi
->s_blocks_per_group
);
2801 if (sbi
->s_inodes_per_group
> blocksize
* 8) {
2802 ext4_msg(sb
, KERN_ERR
,
2803 "#inodes per group too big: %lu",
2804 sbi
->s_inodes_per_group
);
2809 * Test whether we have more sectors than will fit in sector_t,
2810 * and whether the max offset is addressable by the page cache.
2812 if ((ext4_blocks_count(es
) >
2813 (sector_t
)(~0ULL) >> (sb
->s_blocksize_bits
- 9)) ||
2814 (ext4_blocks_count(es
) >
2815 (pgoff_t
)(~0ULL) >> (PAGE_CACHE_SHIFT
- sb
->s_blocksize_bits
))) {
2816 ext4_msg(sb
, KERN_ERR
, "filesystem"
2817 " too large to mount safely on this system");
2818 if (sizeof(sector_t
) < 8)
2819 ext4_msg(sb
, KERN_WARNING
, "CONFIG_LBDAF not enabled");
2824 if (EXT4_BLOCKS_PER_GROUP(sb
) == 0)
2827 /* check blocks count against device size */
2828 blocks_count
= sb
->s_bdev
->bd_inode
->i_size
>> sb
->s_blocksize_bits
;
2829 if (blocks_count
&& ext4_blocks_count(es
) > blocks_count
) {
2830 ext4_msg(sb
, KERN_WARNING
, "bad geometry: block count %llu "
2831 "exceeds size of device (%llu blocks)",
2832 ext4_blocks_count(es
), blocks_count
);
2837 * It makes no sense for the first data block to be beyond the end
2838 * of the filesystem.
2840 if (le32_to_cpu(es
->s_first_data_block
) >= ext4_blocks_count(es
)) {
2841 ext4_msg(sb
, KERN_WARNING
, "bad geometry: first data"
2842 "block %u is beyond end of filesystem (%llu)",
2843 le32_to_cpu(es
->s_first_data_block
),
2844 ext4_blocks_count(es
));
2847 blocks_count
= (ext4_blocks_count(es
) -
2848 le32_to_cpu(es
->s_first_data_block
) +
2849 EXT4_BLOCKS_PER_GROUP(sb
) - 1);
2850 do_div(blocks_count
, EXT4_BLOCKS_PER_GROUP(sb
));
2851 if (blocks_count
> ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb
)) {
2852 ext4_msg(sb
, KERN_WARNING
, "groups count too large: %u "
2853 "(block count %llu, first data block %u, "
2854 "blocks per group %lu)", sbi
->s_groups_count
,
2855 ext4_blocks_count(es
),
2856 le32_to_cpu(es
->s_first_data_block
),
2857 EXT4_BLOCKS_PER_GROUP(sb
));
2860 sbi
->s_groups_count
= blocks_count
;
2861 sbi
->s_blockfile_groups
= min_t(ext4_group_t
, sbi
->s_groups_count
,
2862 (EXT4_MAX_BLOCK_FILE_PHYS
/ EXT4_BLOCKS_PER_GROUP(sb
)));
2863 db_count
= (sbi
->s_groups_count
+ EXT4_DESC_PER_BLOCK(sb
) - 1) /
2864 EXT4_DESC_PER_BLOCK(sb
);
2865 sbi
->s_group_desc
= kmalloc(db_count
* sizeof(struct buffer_head
*),
2867 if (sbi
->s_group_desc
== NULL
) {
2868 ext4_msg(sb
, KERN_ERR
, "not enough memory");
2872 #ifdef CONFIG_PROC_FS
2874 sbi
->s_proc
= proc_mkdir(sb
->s_id
, ext4_proc_root
);
2877 bgl_lock_init(sbi
->s_blockgroup_lock
);
2879 for (i
= 0; i
< db_count
; i
++) {
2880 block
= descriptor_loc(sb
, logical_sb_block
, i
);
2881 sbi
->s_group_desc
[i
] = sb_bread(sb
, block
);
2882 if (!sbi
->s_group_desc
[i
]) {
2883 ext4_msg(sb
, KERN_ERR
,
2884 "can't read group descriptor %d", i
);
2889 if (!ext4_check_descriptors(sb
)) {
2890 ext4_msg(sb
, KERN_ERR
, "group descriptors corrupted!");
2893 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_FLEX_BG
))
2894 if (!ext4_fill_flex_info(sb
)) {
2895 ext4_msg(sb
, KERN_ERR
,
2896 "unable to initialize "
2897 "flex_bg meta info!");
2901 sbi
->s_gdb_count
= db_count
;
2902 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
2903 spin_lock_init(&sbi
->s_next_gen_lock
);
2905 sbi
->s_stripe
= ext4_get_stripe_size(sbi
);
2906 sbi
->s_max_writeback_mb_bump
= 128;
2909 * set up enough so that it can read an inode
2911 if (!test_opt(sb
, NOLOAD
) &&
2912 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
))
2913 sb
->s_op
= &ext4_sops
;
2915 sb
->s_op
= &ext4_nojournal_sops
;
2916 sb
->s_export_op
= &ext4_export_ops
;
2917 sb
->s_xattr
= ext4_xattr_handlers
;
2919 sb
->s_qcop
= &ext4_qctl_operations
;
2920 sb
->dq_op
= &ext4_quota_operations
;
2922 INIT_LIST_HEAD(&sbi
->s_orphan
); /* unlinked but open files */
2923 mutex_init(&sbi
->s_orphan_lock
);
2924 mutex_init(&sbi
->s_resize_lock
);
2928 needs_recovery
= (es
->s_last_orphan
!= 0 ||
2929 EXT4_HAS_INCOMPAT_FEATURE(sb
,
2930 EXT4_FEATURE_INCOMPAT_RECOVER
));
2933 * The first inode we look at is the journal inode. Don't try
2934 * root first: it may be modified in the journal!
2936 if (!test_opt(sb
, NOLOAD
) &&
2937 EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
2938 if (ext4_load_journal(sb
, es
, journal_devnum
))
2940 } else if (test_opt(sb
, NOLOAD
) && !(sb
->s_flags
& MS_RDONLY
) &&
2941 EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
2942 ext4_msg(sb
, KERN_ERR
, "required journal recovery "
2943 "suppressed and not mounted read-only");
2944 goto failed_mount_wq
;
2946 clear_opt(sbi
->s_mount_opt
, DATA_FLAGS
);
2947 set_opt(sbi
->s_mount_opt
, WRITEBACK_DATA
);
2948 sbi
->s_journal
= NULL
;
2953 if (ext4_blocks_count(es
) > 0xffffffffULL
&&
2954 !jbd2_journal_set_features(EXT4_SB(sb
)->s_journal
, 0, 0,
2955 JBD2_FEATURE_INCOMPAT_64BIT
)) {
2956 ext4_msg(sb
, KERN_ERR
, "Failed to set 64-bit journal feature");
2957 goto failed_mount_wq
;
2960 if (test_opt(sb
, JOURNAL_ASYNC_COMMIT
)) {
2961 jbd2_journal_set_features(sbi
->s_journal
,
2962 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2963 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2964 } else if (test_opt(sb
, JOURNAL_CHECKSUM
)) {
2965 jbd2_journal_set_features(sbi
->s_journal
,
2966 JBD2_FEATURE_COMPAT_CHECKSUM
, 0, 0);
2967 jbd2_journal_clear_features(sbi
->s_journal
, 0, 0,
2968 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2970 jbd2_journal_clear_features(sbi
->s_journal
,
2971 JBD2_FEATURE_COMPAT_CHECKSUM
, 0,
2972 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT
);
2975 /* We have now updated the journal if required, so we can
2976 * validate the data journaling mode. */
2977 switch (test_opt(sb
, DATA_FLAGS
)) {
2979 /* No mode set, assume a default based on the journal
2980 * capabilities: ORDERED_DATA if the journal can
2981 * cope, else JOURNAL_DATA
2983 if (jbd2_journal_check_available_features
2984 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
))
2985 set_opt(sbi
->s_mount_opt
, ORDERED_DATA
);
2987 set_opt(sbi
->s_mount_opt
, JOURNAL_DATA
);
2990 case EXT4_MOUNT_ORDERED_DATA
:
2991 case EXT4_MOUNT_WRITEBACK_DATA
:
2992 if (!jbd2_journal_check_available_features
2993 (sbi
->s_journal
, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE
)) {
2994 ext4_msg(sb
, KERN_ERR
, "Journal does not support "
2995 "requested data journaling mode");
2996 goto failed_mount_wq
;
3001 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3004 err
= percpu_counter_init(&sbi
->s_freeblocks_counter
,
3005 ext4_count_free_blocks(sb
));
3007 err
= percpu_counter_init(&sbi
->s_freeinodes_counter
,
3008 ext4_count_free_inodes(sb
));
3010 err
= percpu_counter_init(&sbi
->s_dirs_counter
,
3011 ext4_count_dirs(sb
));
3013 err
= percpu_counter_init(&sbi
->s_dirtyblocks_counter
, 0);
3015 ext4_msg(sb
, KERN_ERR
, "insufficient memory");
3016 goto failed_mount_wq
;
3019 EXT4_SB(sb
)->dio_unwritten_wq
= create_workqueue("ext4-dio-unwritten");
3020 if (!EXT4_SB(sb
)->dio_unwritten_wq
) {
3021 printk(KERN_ERR
"EXT4-fs: failed to create DIO workqueue\n");
3022 goto failed_mount_wq
;
3026 * The jbd2_journal_load will have done any necessary log recovery,
3027 * so we can safely mount the rest of the filesystem now.
3030 root
= ext4_iget(sb
, EXT4_ROOT_INO
);
3032 ext4_msg(sb
, KERN_ERR
, "get root inode failed");
3033 ret
= PTR_ERR(root
);
3036 if (!S_ISDIR(root
->i_mode
) || !root
->i_blocks
|| !root
->i_size
) {
3038 ext4_msg(sb
, KERN_ERR
, "corrupt root inode, run e2fsck");
3041 sb
->s_root
= d_alloc_root(root
);
3043 ext4_msg(sb
, KERN_ERR
, "get root dentry failed");
3049 ext4_setup_super(sb
, es
, sb
->s_flags
& MS_RDONLY
);
3051 /* determine the minimum size of new large inodes, if present */
3052 if (sbi
->s_inode_size
> EXT4_GOOD_OLD_INODE_SIZE
) {
3053 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3054 EXT4_GOOD_OLD_INODE_SIZE
;
3055 if (EXT4_HAS_RO_COMPAT_FEATURE(sb
,
3056 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE
)) {
3057 if (sbi
->s_want_extra_isize
<
3058 le16_to_cpu(es
->s_want_extra_isize
))
3059 sbi
->s_want_extra_isize
=
3060 le16_to_cpu(es
->s_want_extra_isize
);
3061 if (sbi
->s_want_extra_isize
<
3062 le16_to_cpu(es
->s_min_extra_isize
))
3063 sbi
->s_want_extra_isize
=
3064 le16_to_cpu(es
->s_min_extra_isize
);
3067 /* Check if enough inode space is available */
3068 if (EXT4_GOOD_OLD_INODE_SIZE
+ sbi
->s_want_extra_isize
>
3069 sbi
->s_inode_size
) {
3070 sbi
->s_want_extra_isize
= sizeof(struct ext4_inode
) -
3071 EXT4_GOOD_OLD_INODE_SIZE
;
3072 ext4_msg(sb
, KERN_INFO
, "required extra inode space not"
3076 if (test_opt(sb
, DELALLOC
) &&
3077 (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)) {
3078 ext4_msg(sb
, KERN_WARNING
, "Ignoring delalloc option - "
3079 "requested data journaling mode");
3080 clear_opt(sbi
->s_mount_opt
, DELALLOC
);
3082 if (test_opt(sb
, DIOREAD_NOLOCK
)) {
3083 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
) {
3084 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
3085 "option - requested data journaling mode");
3086 clear_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
3088 if (sb
->s_blocksize
< PAGE_SIZE
) {
3089 ext4_msg(sb
, KERN_WARNING
, "Ignoring dioread_nolock "
3090 "option - block size is too small");
3091 clear_opt(sbi
->s_mount_opt
, DIOREAD_NOLOCK
);
3095 err
= ext4_setup_system_zone(sb
);
3097 ext4_msg(sb
, KERN_ERR
, "failed to initialize system "
3103 err
= ext4_mb_init(sb
, needs_recovery
);
3105 ext4_msg(sb
, KERN_ERR
, "failed to initalize mballoc (%d)",
3110 sbi
->s_kobj
.kset
= ext4_kset
;
3111 init_completion(&sbi
->s_kobj_unregister
);
3112 err
= kobject_init_and_add(&sbi
->s_kobj
, &ext4_ktype
, NULL
,
3115 ext4_mb_release(sb
);
3116 ext4_ext_release(sb
);
3120 EXT4_SB(sb
)->s_mount_state
|= EXT4_ORPHAN_FS
;
3121 ext4_orphan_cleanup(sb
, es
);
3122 EXT4_SB(sb
)->s_mount_state
&= ~EXT4_ORPHAN_FS
;
3123 if (needs_recovery
) {
3124 ext4_msg(sb
, KERN_INFO
, "recovery complete");
3125 ext4_mark_recovery_complete(sb
, es
);
3127 if (EXT4_SB(sb
)->s_journal
) {
3128 if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
3129 descr
= " journalled data mode";
3130 else if (test_opt(sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
3131 descr
= " ordered data mode";
3133 descr
= " writeback data mode";
3135 descr
= "out journal";
3137 ext4_msg(sb
, KERN_INFO
, "mounted filesystem with%s. "
3138 "Opts: %s", descr
, orig_data
);
3140 init_timer(&sbi
->s_err_report
);
3141 sbi
->s_err_report
.function
= print_daily_error_info
;
3142 sbi
->s_err_report
.data
= (unsigned long) sb
;
3143 if (es
->s_error_count
)
3144 mod_timer(&sbi
->s_err_report
, jiffies
+ 300*HZ
); /* 5 minutes */
3152 ext4_msg(sb
, KERN_ERR
, "VFS: Can't find ext4 filesystem");
3156 ext4_msg(sb
, KERN_ERR
, "mount failed");
3157 destroy_workqueue(EXT4_SB(sb
)->dio_unwritten_wq
);
3159 ext4_release_system_zone(sb
);
3160 if (sbi
->s_journal
) {
3161 jbd2_journal_destroy(sbi
->s_journal
);
3162 sbi
->s_journal
= NULL
;
3164 percpu_counter_destroy(&sbi
->s_freeblocks_counter
);
3165 percpu_counter_destroy(&sbi
->s_freeinodes_counter
);
3166 percpu_counter_destroy(&sbi
->s_dirs_counter
);
3167 percpu_counter_destroy(&sbi
->s_dirtyblocks_counter
);
3169 if (sbi
->s_flex_groups
) {
3170 if (is_vmalloc_addr(sbi
->s_flex_groups
))
3171 vfree(sbi
->s_flex_groups
);
3173 kfree(sbi
->s_flex_groups
);
3176 for (i
= 0; i
< db_count
; i
++)
3177 brelse(sbi
->s_group_desc
[i
]);
3178 kfree(sbi
->s_group_desc
);
3181 remove_proc_entry(sb
->s_id
, ext4_proc_root
);
3184 for (i
= 0; i
< MAXQUOTAS
; i
++)
3185 kfree(sbi
->s_qf_names
[i
]);
3187 ext4_blkdev_remove(sbi
);
3190 sb
->s_fs_info
= NULL
;
3191 kfree(sbi
->s_blockgroup_lock
);
3200 * Setup any per-fs journal parameters now. We'll do this both on
3201 * initial mount, once the journal has been initialised but before we've
3202 * done any recovery; and again on any subsequent remount.
3204 static void ext4_init_journal_params(struct super_block
*sb
, journal_t
*journal
)
3206 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3208 journal
->j_commit_interval
= sbi
->s_commit_interval
;
3209 journal
->j_min_batch_time
= sbi
->s_min_batch_time
;
3210 journal
->j_max_batch_time
= sbi
->s_max_batch_time
;
3212 spin_lock(&journal
->j_state_lock
);
3213 if (test_opt(sb
, BARRIER
))
3214 journal
->j_flags
|= JBD2_BARRIER
;
3216 journal
->j_flags
&= ~JBD2_BARRIER
;
3217 if (test_opt(sb
, DATA_ERR_ABORT
))
3218 journal
->j_flags
|= JBD2_ABORT_ON_SYNCDATA_ERR
;
3220 journal
->j_flags
&= ~JBD2_ABORT_ON_SYNCDATA_ERR
;
3221 spin_unlock(&journal
->j_state_lock
);
3224 static journal_t
*ext4_get_journal(struct super_block
*sb
,
3225 unsigned int journal_inum
)
3227 struct inode
*journal_inode
;
3230 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3232 /* First, test for the existence of a valid inode on disk. Bad
3233 * things happen if we iget() an unused inode, as the subsequent
3234 * iput() will try to delete it. */
3236 journal_inode
= ext4_iget(sb
, journal_inum
);
3237 if (IS_ERR(journal_inode
)) {
3238 ext4_msg(sb
, KERN_ERR
, "no journal found");
3241 if (!journal_inode
->i_nlink
) {
3242 make_bad_inode(journal_inode
);
3243 iput(journal_inode
);
3244 ext4_msg(sb
, KERN_ERR
, "journal inode is deleted");
3248 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3249 journal_inode
, journal_inode
->i_size
);
3250 if (!S_ISREG(journal_inode
->i_mode
)) {
3251 ext4_msg(sb
, KERN_ERR
, "invalid journal inode");
3252 iput(journal_inode
);
3256 journal
= jbd2_journal_init_inode(journal_inode
);
3258 ext4_msg(sb
, KERN_ERR
, "Could not load journal inode");
3259 iput(journal_inode
);
3262 journal
->j_private
= sb
;
3263 ext4_init_journal_params(sb
, journal
);
3267 static journal_t
*ext4_get_dev_journal(struct super_block
*sb
,
3270 struct buffer_head
*bh
;
3274 int hblock
, blocksize
;
3275 ext4_fsblk_t sb_block
;
3276 unsigned long offset
;
3277 struct ext4_super_block
*es
;
3278 struct block_device
*bdev
;
3280 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3282 bdev
= ext4_blkdev_get(j_dev
, sb
);
3286 if (bd_claim(bdev
, sb
)) {
3287 ext4_msg(sb
, KERN_ERR
,
3288 "failed to claim external journal device");
3289 blkdev_put(bdev
, FMODE_READ
|FMODE_WRITE
);
3293 blocksize
= sb
->s_blocksize
;
3294 hblock
= bdev_logical_block_size(bdev
);
3295 if (blocksize
< hblock
) {
3296 ext4_msg(sb
, KERN_ERR
,
3297 "blocksize too small for journal device");
3301 sb_block
= EXT4_MIN_BLOCK_SIZE
/ blocksize
;
3302 offset
= EXT4_MIN_BLOCK_SIZE
% blocksize
;
3303 set_blocksize(bdev
, blocksize
);
3304 if (!(bh
= __bread(bdev
, sb_block
, blocksize
))) {
3305 ext4_msg(sb
, KERN_ERR
, "couldn't read superblock of "
3306 "external journal");
3310 es
= (struct ext4_super_block
*) (((char *)bh
->b_data
) + offset
);
3311 if ((le16_to_cpu(es
->s_magic
) != EXT4_SUPER_MAGIC
) ||
3312 !(le32_to_cpu(es
->s_feature_incompat
) &
3313 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV
)) {
3314 ext4_msg(sb
, KERN_ERR
, "external journal has "
3320 if (memcmp(EXT4_SB(sb
)->s_es
->s_journal_uuid
, es
->s_uuid
, 16)) {
3321 ext4_msg(sb
, KERN_ERR
, "journal UUID does not match");
3326 len
= ext4_blocks_count(es
);
3327 start
= sb_block
+ 1;
3328 brelse(bh
); /* we're done with the superblock */
3330 journal
= jbd2_journal_init_dev(bdev
, sb
->s_bdev
,
3331 start
, len
, blocksize
);
3333 ext4_msg(sb
, KERN_ERR
, "failed to create device journal");
3336 journal
->j_private
= sb
;
3337 ll_rw_block(READ
, 1, &journal
->j_sb_buffer
);
3338 wait_on_buffer(journal
->j_sb_buffer
);
3339 if (!buffer_uptodate(journal
->j_sb_buffer
)) {
3340 ext4_msg(sb
, KERN_ERR
, "I/O error on journal device");
3343 if (be32_to_cpu(journal
->j_superblock
->s_nr_users
) != 1) {
3344 ext4_msg(sb
, KERN_ERR
, "External journal has more than one "
3345 "user (unsupported) - %d",
3346 be32_to_cpu(journal
->j_superblock
->s_nr_users
));
3349 EXT4_SB(sb
)->journal_bdev
= bdev
;
3350 ext4_init_journal_params(sb
, journal
);
3354 jbd2_journal_destroy(journal
);
3356 ext4_blkdev_put(bdev
);
3360 static int ext4_load_journal(struct super_block
*sb
,
3361 struct ext4_super_block
*es
,
3362 unsigned long journal_devnum
)
3365 unsigned int journal_inum
= le32_to_cpu(es
->s_journal_inum
);
3368 int really_read_only
;
3370 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3372 if (journal_devnum
&&
3373 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3374 ext4_msg(sb
, KERN_INFO
, "external journal device major/minor "
3375 "numbers have changed");
3376 journal_dev
= new_decode_dev(journal_devnum
);
3378 journal_dev
= new_decode_dev(le32_to_cpu(es
->s_journal_dev
));
3380 really_read_only
= bdev_read_only(sb
->s_bdev
);
3383 * Are we loading a blank journal or performing recovery after a
3384 * crash? For recovery, we need to check in advance whether we
3385 * can get read-write access to the device.
3387 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
)) {
3388 if (sb
->s_flags
& MS_RDONLY
) {
3389 ext4_msg(sb
, KERN_INFO
, "INFO: recovery "
3390 "required on readonly filesystem");
3391 if (really_read_only
) {
3392 ext4_msg(sb
, KERN_ERR
, "write access "
3393 "unavailable, cannot proceed");
3396 ext4_msg(sb
, KERN_INFO
, "write access will "
3397 "be enabled during recovery");
3401 if (journal_inum
&& journal_dev
) {
3402 ext4_msg(sb
, KERN_ERR
, "filesystem has both journal "
3403 "and inode journals!");
3408 if (!(journal
= ext4_get_journal(sb
, journal_inum
)))
3411 if (!(journal
= ext4_get_dev_journal(sb
, journal_dev
)))
3415 if (!(journal
->j_flags
& JBD2_BARRIER
))
3416 ext4_msg(sb
, KERN_INFO
, "barriers disabled");
3418 if (!really_read_only
&& test_opt(sb
, UPDATE_JOURNAL
)) {
3419 err
= jbd2_journal_update_format(journal
);
3421 ext4_msg(sb
, KERN_ERR
, "error updating journal");
3422 jbd2_journal_destroy(journal
);
3427 if (!EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
))
3428 err
= jbd2_journal_wipe(journal
, !really_read_only
);
3430 char *save
= kmalloc(EXT4_S_ERR_LEN
, GFP_KERNEL
);
3432 memcpy(save
, ((char *) es
) +
3433 EXT4_S_ERR_START
, EXT4_S_ERR_LEN
);
3434 err
= jbd2_journal_load(journal
);
3436 memcpy(((char *) es
) + EXT4_S_ERR_START
,
3437 save
, EXT4_S_ERR_LEN
);
3442 ext4_msg(sb
, KERN_ERR
, "error loading journal");
3443 jbd2_journal_destroy(journal
);
3447 EXT4_SB(sb
)->s_journal
= journal
;
3448 ext4_clear_journal_err(sb
, es
);
3450 if (journal_devnum
&&
3451 journal_devnum
!= le32_to_cpu(es
->s_journal_dev
)) {
3452 es
->s_journal_dev
= cpu_to_le32(journal_devnum
);
3454 /* Make sure we flush the recovery flag to disk. */
3455 ext4_commit_super(sb
, 1);
3461 static int ext4_commit_super(struct super_block
*sb
, int sync
)
3463 struct ext4_super_block
*es
= EXT4_SB(sb
)->s_es
;
3464 struct buffer_head
*sbh
= EXT4_SB(sb
)->s_sbh
;
3469 if (buffer_write_io_error(sbh
)) {
3471 * Oh, dear. A previous attempt to write the
3472 * superblock failed. This could happen because the
3473 * USB device was yanked out. Or it could happen to
3474 * be a transient write error and maybe the block will
3475 * be remapped. Nothing we can do but to retry the
3476 * write and hope for the best.
3478 ext4_msg(sb
, KERN_ERR
, "previous I/O error to "
3479 "superblock detected");
3480 clear_buffer_write_io_error(sbh
);
3481 set_buffer_uptodate(sbh
);
3484 * If the file system is mounted read-only, don't update the
3485 * superblock write time. This avoids updating the superblock
3486 * write time when we are mounting the root file system
3487 * read/only but we need to replay the journal; at that point,
3488 * for people who are east of GMT and who make their clock
3489 * tick in localtime for Windows bug-for-bug compatibility,
3490 * the clock is set in the future, and this will cause e2fsck
3491 * to complain and force a full file system check.
3493 if (!(sb
->s_flags
& MS_RDONLY
))
3494 es
->s_wtime
= cpu_to_le32(get_seconds());
3495 es
->s_kbytes_written
=
3496 cpu_to_le64(EXT4_SB(sb
)->s_kbytes_written
+
3497 ((part_stat_read(sb
->s_bdev
->bd_part
, sectors
[1]) -
3498 EXT4_SB(sb
)->s_sectors_written_start
) >> 1));
3499 ext4_free_blocks_count_set(es
, percpu_counter_sum_positive(
3500 &EXT4_SB(sb
)->s_freeblocks_counter
));
3501 es
->s_free_inodes_count
= cpu_to_le32(percpu_counter_sum_positive(
3502 &EXT4_SB(sb
)->s_freeinodes_counter
));
3504 BUFFER_TRACE(sbh
, "marking dirty");
3505 mark_buffer_dirty(sbh
);
3507 error
= sync_dirty_buffer(sbh
);
3511 error
= buffer_write_io_error(sbh
);
3513 ext4_msg(sb
, KERN_ERR
, "I/O error while writing "
3515 clear_buffer_write_io_error(sbh
);
3516 set_buffer_uptodate(sbh
);
3523 * Have we just finished recovery? If so, and if we are mounting (or
3524 * remounting) the filesystem readonly, then we will end up with a
3525 * consistent fs on disk. Record that fact.
3527 static void ext4_mark_recovery_complete(struct super_block
*sb
,
3528 struct ext4_super_block
*es
)
3530 journal_t
*journal
= EXT4_SB(sb
)->s_journal
;
3532 if (!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
)) {
3533 BUG_ON(journal
!= NULL
);
3536 jbd2_journal_lock_updates(journal
);
3537 if (jbd2_journal_flush(journal
) < 0)
3540 if (EXT4_HAS_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
) &&
3541 sb
->s_flags
& MS_RDONLY
) {
3542 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3543 ext4_commit_super(sb
, 1);
3547 jbd2_journal_unlock_updates(journal
);
3551 * If we are mounting (or read-write remounting) a filesystem whose journal
3552 * has recorded an error from a previous lifetime, move that error to the
3553 * main filesystem now.
3555 static void ext4_clear_journal_err(struct super_block
*sb
,
3556 struct ext4_super_block
*es
)
3562 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb
, EXT4_FEATURE_COMPAT_HAS_JOURNAL
));
3564 journal
= EXT4_SB(sb
)->s_journal
;
3567 * Now check for any error status which may have been recorded in the
3568 * journal by a prior ext4_error() or ext4_abort()
3571 j_errno
= jbd2_journal_errno(journal
);
3575 errstr
= ext4_decode_error(sb
, j_errno
, nbuf
);
3576 ext4_warning(sb
, "Filesystem error recorded "
3577 "from previous mount: %s", errstr
);
3578 ext4_warning(sb
, "Marking fs in need of filesystem check.");
3580 EXT4_SB(sb
)->s_mount_state
|= EXT4_ERROR_FS
;
3581 es
->s_state
|= cpu_to_le16(EXT4_ERROR_FS
);
3582 ext4_commit_super(sb
, 1);
3584 jbd2_journal_clear_err(journal
);
3589 * Force the running and committing transactions to commit,
3590 * and wait on the commit.
3592 int ext4_force_commit(struct super_block
*sb
)
3597 if (sb
->s_flags
& MS_RDONLY
)
3600 journal
= EXT4_SB(sb
)->s_journal
;
3602 vfs_check_frozen(sb
, SB_FREEZE_WRITE
);
3603 ret
= ext4_journal_force_commit(journal
);
3609 static void ext4_write_super(struct super_block
*sb
)
3612 ext4_commit_super(sb
, 1);
3616 static int ext4_sync_fs(struct super_block
*sb
, int wait
)
3620 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3622 trace_ext4_sync_fs(sb
, wait
);
3623 flush_workqueue(sbi
->dio_unwritten_wq
);
3624 if (jbd2_journal_start_commit(sbi
->s_journal
, &target
)) {
3626 jbd2_log_wait_commit(sbi
->s_journal
, target
);
3632 * LVM calls this function before a (read-only) snapshot is created. This
3633 * gives us a chance to flush the journal completely and mark the fs clean.
3635 static int ext4_freeze(struct super_block
*sb
)
3640 if (sb
->s_flags
& MS_RDONLY
)
3643 journal
= EXT4_SB(sb
)->s_journal
;
3645 /* Now we set up the journal barrier. */
3646 jbd2_journal_lock_updates(journal
);
3649 * Don't clear the needs_recovery flag if we failed to flush
3652 error
= jbd2_journal_flush(journal
);
3656 /* Journal blocked and flushed, clear needs_recovery flag. */
3657 EXT4_CLEAR_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3658 error
= ext4_commit_super(sb
, 1);
3660 /* we rely on s_frozen to stop further updates */
3661 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
3666 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3667 * flag here, even though the filesystem is not technically dirty yet.
3669 static int ext4_unfreeze(struct super_block
*sb
)
3671 if (sb
->s_flags
& MS_RDONLY
)
3675 /* Reset the needs_recovery flag before the fs is unlocked. */
3676 EXT4_SET_INCOMPAT_FEATURE(sb
, EXT4_FEATURE_INCOMPAT_RECOVER
);
3677 ext4_commit_super(sb
, 1);
3682 static int ext4_remount(struct super_block
*sb
, int *flags
, char *data
)
3684 struct ext4_super_block
*es
;
3685 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3686 ext4_fsblk_t n_blocks_count
= 0;
3687 unsigned long old_sb_flags
;
3688 struct ext4_mount_options old_opts
;
3689 int enable_quota
= 0;
3691 unsigned int journal_ioprio
= DEFAULT_JOURNAL_IOPRIO
;
3696 char *orig_data
= kstrdup(data
, GFP_KERNEL
);
3700 /* Store the original options */
3702 old_sb_flags
= sb
->s_flags
;
3703 old_opts
.s_mount_opt
= sbi
->s_mount_opt
;
3704 old_opts
.s_resuid
= sbi
->s_resuid
;
3705 old_opts
.s_resgid
= sbi
->s_resgid
;
3706 old_opts
.s_commit_interval
= sbi
->s_commit_interval
;
3707 old_opts
.s_min_batch_time
= sbi
->s_min_batch_time
;
3708 old_opts
.s_max_batch_time
= sbi
->s_max_batch_time
;
3710 old_opts
.s_jquota_fmt
= sbi
->s_jquota_fmt
;
3711 for (i
= 0; i
< MAXQUOTAS
; i
++)
3712 old_opts
.s_qf_names
[i
] = sbi
->s_qf_names
[i
];
3714 if (sbi
->s_journal
&& sbi
->s_journal
->j_task
->io_context
)
3715 journal_ioprio
= sbi
->s_journal
->j_task
->io_context
->ioprio
;
3718 * Allow the "check" option to be passed as a remount option.
3720 if (!parse_options(data
, sb
, NULL
, &journal_ioprio
,
3721 &n_blocks_count
, 1)) {
3726 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
)
3727 ext4_abort(sb
, "Abort forced by user");
3729 sb
->s_flags
= (sb
->s_flags
& ~MS_POSIXACL
) |
3730 (test_opt(sb
, POSIX_ACL
) ? MS_POSIXACL
: 0);
3734 if (sbi
->s_journal
) {
3735 ext4_init_journal_params(sb
, sbi
->s_journal
);
3736 set_task_ioprio(sbi
->s_journal
->j_task
, journal_ioprio
);
3739 if ((*flags
& MS_RDONLY
) != (sb
->s_flags
& MS_RDONLY
) ||
3740 n_blocks_count
> ext4_blocks_count(es
)) {
3741 if (sbi
->s_mount_flags
& EXT4_MF_FS_ABORTED
) {
3746 if (*flags
& MS_RDONLY
) {
3747 err
= dquot_suspend(sb
, -1);
3752 * First of all, the unconditional stuff we have to do
3753 * to disable replay of the journal when we next remount
3755 sb
->s_flags
|= MS_RDONLY
;
3758 * OK, test if we are remounting a valid rw partition
3759 * readonly, and if so set the rdonly flag and then
3760 * mark the partition as valid again.
3762 if (!(es
->s_state
& cpu_to_le16(EXT4_VALID_FS
)) &&
3763 (sbi
->s_mount_state
& EXT4_VALID_FS
))
3764 es
->s_state
= cpu_to_le16(sbi
->s_mount_state
);
3767 ext4_mark_recovery_complete(sb
, es
);
3769 /* Make sure we can mount this feature set readwrite */
3770 if (!ext4_feature_set_ok(sb
, 0)) {
3775 * Make sure the group descriptor checksums
3776 * are sane. If they aren't, refuse to remount r/w.
3778 for (g
= 0; g
< sbi
->s_groups_count
; g
++) {
3779 struct ext4_group_desc
*gdp
=
3780 ext4_get_group_desc(sb
, g
, NULL
);
3782 if (!ext4_group_desc_csum_verify(sbi
, g
, gdp
)) {
3783 ext4_msg(sb
, KERN_ERR
,
3784 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3785 g
, le16_to_cpu(ext4_group_desc_csum(sbi
, g
, gdp
)),
3786 le16_to_cpu(gdp
->bg_checksum
));
3793 * If we have an unprocessed orphan list hanging
3794 * around from a previously readonly bdev mount,
3795 * require a full umount/remount for now.
3797 if (es
->s_last_orphan
) {
3798 ext4_msg(sb
, KERN_WARNING
, "Couldn't "
3799 "remount RDWR because of unprocessed "
3800 "orphan inode list. Please "
3801 "umount/remount instead");
3807 * Mounting a RDONLY partition read-write, so reread
3808 * and store the current valid flag. (It may have
3809 * been changed by e2fsck since we originally mounted
3813 ext4_clear_journal_err(sb
, es
);
3814 sbi
->s_mount_state
= le16_to_cpu(es
->s_state
);
3815 if ((err
= ext4_group_extend(sb
, es
, n_blocks_count
)))
3817 if (!ext4_setup_super(sb
, es
, 0))
3818 sb
->s_flags
&= ~MS_RDONLY
;
3822 ext4_setup_system_zone(sb
);
3823 if (sbi
->s_journal
== NULL
)
3824 ext4_commit_super(sb
, 1);
3827 /* Release old quota file names */
3828 for (i
= 0; i
< MAXQUOTAS
; i
++)
3829 if (old_opts
.s_qf_names
[i
] &&
3830 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3831 kfree(old_opts
.s_qf_names
[i
]);
3836 dquot_resume(sb
, -1);
3838 ext4_msg(sb
, KERN_INFO
, "re-mounted. Opts: %s", orig_data
);
3843 sb
->s_flags
= old_sb_flags
;
3844 sbi
->s_mount_opt
= old_opts
.s_mount_opt
;
3845 sbi
->s_resuid
= old_opts
.s_resuid
;
3846 sbi
->s_resgid
= old_opts
.s_resgid
;
3847 sbi
->s_commit_interval
= old_opts
.s_commit_interval
;
3848 sbi
->s_min_batch_time
= old_opts
.s_min_batch_time
;
3849 sbi
->s_max_batch_time
= old_opts
.s_max_batch_time
;
3851 sbi
->s_jquota_fmt
= old_opts
.s_jquota_fmt
;
3852 for (i
= 0; i
< MAXQUOTAS
; i
++) {
3853 if (sbi
->s_qf_names
[i
] &&
3854 old_opts
.s_qf_names
[i
] != sbi
->s_qf_names
[i
])
3855 kfree(sbi
->s_qf_names
[i
]);
3856 sbi
->s_qf_names
[i
] = old_opts
.s_qf_names
[i
];
3865 static int ext4_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
3867 struct super_block
*sb
= dentry
->d_sb
;
3868 struct ext4_sb_info
*sbi
= EXT4_SB(sb
);
3869 struct ext4_super_block
*es
= sbi
->s_es
;
3872 if (test_opt(sb
, MINIX_DF
)) {
3873 sbi
->s_overhead_last
= 0;
3874 } else if (sbi
->s_blocks_last
!= ext4_blocks_count(es
)) {
3875 ext4_group_t i
, ngroups
= ext4_get_groups_count(sb
);
3876 ext4_fsblk_t overhead
= 0;
3879 * Compute the overhead (FS structures). This is constant
3880 * for a given filesystem unless the number of block groups
3881 * changes so we cache the previous value until it does.
3885 * All of the blocks before first_data_block are
3888 overhead
= le32_to_cpu(es
->s_first_data_block
);
3891 * Add the overhead attributed to the superblock and
3892 * block group descriptors. If the sparse superblocks
3893 * feature is turned on, then not all groups have this.
3895 for (i
= 0; i
< ngroups
; i
++) {
3896 overhead
+= ext4_bg_has_super(sb
, i
) +
3897 ext4_bg_num_gdb(sb
, i
);
3902 * Every block group has an inode bitmap, a block
3903 * bitmap, and an inode table.
3905 overhead
+= ngroups
* (2 + sbi
->s_itb_per_group
);
3906 sbi
->s_overhead_last
= overhead
;
3908 sbi
->s_blocks_last
= ext4_blocks_count(es
);
3911 buf
->f_type
= EXT4_SUPER_MAGIC
;
3912 buf
->f_bsize
= sb
->s_blocksize
;
3913 buf
->f_blocks
= ext4_blocks_count(es
) - sbi
->s_overhead_last
;
3914 buf
->f_bfree
= percpu_counter_sum_positive(&sbi
->s_freeblocks_counter
) -
3915 percpu_counter_sum_positive(&sbi
->s_dirtyblocks_counter
);
3916 buf
->f_bavail
= buf
->f_bfree
- ext4_r_blocks_count(es
);
3917 if (buf
->f_bfree
< ext4_r_blocks_count(es
))
3919 buf
->f_files
= le32_to_cpu(es
->s_inodes_count
);
3920 buf
->f_ffree
= percpu_counter_sum_positive(&sbi
->s_freeinodes_counter
);
3921 buf
->f_namelen
= EXT4_NAME_LEN
;
3922 fsid
= le64_to_cpup((void *)es
->s_uuid
) ^
3923 le64_to_cpup((void *)es
->s_uuid
+ sizeof(u64
));
3924 buf
->f_fsid
.val
[0] = fsid
& 0xFFFFFFFFUL
;
3925 buf
->f_fsid
.val
[1] = (fsid
>> 32) & 0xFFFFFFFFUL
;
3930 /* Helper function for writing quotas on sync - we need to start transaction
3931 * before quota file is locked for write. Otherwise the are possible deadlocks:
3932 * Process 1 Process 2
3933 * ext4_create() quota_sync()
3934 * jbd2_journal_start() write_dquot()
3935 * dquot_initialize() down(dqio_mutex)
3936 * down(dqio_mutex) jbd2_journal_start()
3942 static inline struct inode
*dquot_to_inode(struct dquot
*dquot
)
3944 return sb_dqopt(dquot
->dq_sb
)->files
[dquot
->dq_type
];
3947 static int ext4_write_dquot(struct dquot
*dquot
)
3951 struct inode
*inode
;
3953 inode
= dquot_to_inode(dquot
);
3954 handle
= ext4_journal_start(inode
,
3955 EXT4_QUOTA_TRANS_BLOCKS(dquot
->dq_sb
));
3957 return PTR_ERR(handle
);
3958 ret
= dquot_commit(dquot
);
3959 err
= ext4_journal_stop(handle
);
3965 static int ext4_acquire_dquot(struct dquot
*dquot
)
3970 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3971 EXT4_QUOTA_INIT_BLOCKS(dquot
->dq_sb
));
3973 return PTR_ERR(handle
);
3974 ret
= dquot_acquire(dquot
);
3975 err
= ext4_journal_stop(handle
);
3981 static int ext4_release_dquot(struct dquot
*dquot
)
3986 handle
= ext4_journal_start(dquot_to_inode(dquot
),
3987 EXT4_QUOTA_DEL_BLOCKS(dquot
->dq_sb
));
3988 if (IS_ERR(handle
)) {
3989 /* Release dquot anyway to avoid endless cycle in dqput() */
3990 dquot_release(dquot
);
3991 return PTR_ERR(handle
);
3993 ret
= dquot_release(dquot
);
3994 err
= ext4_journal_stop(handle
);
4000 static int ext4_mark_dquot_dirty(struct dquot
*dquot
)
4002 /* Are we journaling quotas? */
4003 if (EXT4_SB(dquot
->dq_sb
)->s_qf_names
[USRQUOTA
] ||
4004 EXT4_SB(dquot
->dq_sb
)->s_qf_names
[GRPQUOTA
]) {
4005 dquot_mark_dquot_dirty(dquot
);
4006 return ext4_write_dquot(dquot
);
4008 return dquot_mark_dquot_dirty(dquot
);
4012 static int ext4_write_info(struct super_block
*sb
, int type
)
4017 /* Data block + inode block */
4018 handle
= ext4_journal_start(sb
->s_root
->d_inode
, 2);
4020 return PTR_ERR(handle
);
4021 ret
= dquot_commit_info(sb
, type
);
4022 err
= ext4_journal_stop(handle
);
4029 * Turn on quotas during mount time - we need to find
4030 * the quota file and such...
4032 static int ext4_quota_on_mount(struct super_block
*sb
, int type
)
4034 return dquot_quota_on_mount(sb
, EXT4_SB(sb
)->s_qf_names
[type
],
4035 EXT4_SB(sb
)->s_jquota_fmt
, type
);
4039 * Standard function to be called on quota_on
4041 static int ext4_quota_on(struct super_block
*sb
, int type
, int format_id
,
4047 if (!test_opt(sb
, QUOTA
))
4050 err
= kern_path(name
, LOOKUP_FOLLOW
, &path
);
4054 /* Quotafile not on the same filesystem? */
4055 if (path
.mnt
->mnt_sb
!= sb
) {
4059 /* Journaling quota? */
4060 if (EXT4_SB(sb
)->s_qf_names
[type
]) {
4061 /* Quotafile not in fs root? */
4062 if (path
.dentry
->d_parent
!= sb
->s_root
)
4063 ext4_msg(sb
, KERN_WARNING
,
4064 "Quota file not on filesystem root. "
4065 "Journaled quota will not work");
4069 * When we journal data on quota file, we have to flush journal to see
4070 * all updates to the file when we bypass pagecache...
4072 if (EXT4_SB(sb
)->s_journal
&&
4073 ext4_should_journal_data(path
.dentry
->d_inode
)) {
4075 * We don't need to lock updates but journal_flush() could
4076 * otherwise be livelocked...
4078 jbd2_journal_lock_updates(EXT4_SB(sb
)->s_journal
);
4079 err
= jbd2_journal_flush(EXT4_SB(sb
)->s_journal
);
4080 jbd2_journal_unlock_updates(EXT4_SB(sb
)->s_journal
);
4087 err
= dquot_quota_on_path(sb
, type
, format_id
, &path
);
4092 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4093 * acquiring the locks... As quota files are never truncated and quota code
4094 * itself serializes the operations (and noone else should touch the files)
4095 * we don't have to be afraid of races */
4096 static ssize_t
ext4_quota_read(struct super_block
*sb
, int type
, char *data
,
4097 size_t len
, loff_t off
)
4099 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4100 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4102 int offset
= off
& (sb
->s_blocksize
- 1);
4105 struct buffer_head
*bh
;
4106 loff_t i_size
= i_size_read(inode
);
4110 if (off
+len
> i_size
)
4113 while (toread
> 0) {
4114 tocopy
= sb
->s_blocksize
- offset
< toread
?
4115 sb
->s_blocksize
- offset
: toread
;
4116 bh
= ext4_bread(NULL
, inode
, blk
, 0, &err
);
4119 if (!bh
) /* A hole? */
4120 memset(data
, 0, tocopy
);
4122 memcpy(data
, bh
->b_data
+offset
, tocopy
);
4132 /* Write to quotafile (we know the transaction is already started and has
4133 * enough credits) */
4134 static ssize_t
ext4_quota_write(struct super_block
*sb
, int type
,
4135 const char *data
, size_t len
, loff_t off
)
4137 struct inode
*inode
= sb_dqopt(sb
)->files
[type
];
4138 ext4_lblk_t blk
= off
>> EXT4_BLOCK_SIZE_BITS(sb
);
4140 int offset
= off
& (sb
->s_blocksize
- 1);
4141 int journal_quota
= EXT4_SB(sb
)->s_qf_names
[type
] != NULL
;
4142 struct buffer_head
*bh
;
4143 handle_t
*handle
= journal_current_handle();
4145 if (EXT4_SB(sb
)->s_journal
&& !handle
) {
4146 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4147 " cancelled because transaction is not started",
4148 (unsigned long long)off
, (unsigned long long)len
);
4152 * Since we account only one data block in transaction credits,
4153 * then it is impossible to cross a block boundary.
4155 if (sb
->s_blocksize
- offset
< len
) {
4156 ext4_msg(sb
, KERN_WARNING
, "Quota write (off=%llu, len=%llu)"
4157 " cancelled because not block aligned",
4158 (unsigned long long)off
, (unsigned long long)len
);
4162 mutex_lock_nested(&inode
->i_mutex
, I_MUTEX_QUOTA
);
4163 bh
= ext4_bread(handle
, inode
, blk
, 1, &err
);
4166 if (journal_quota
) {
4167 err
= ext4_journal_get_write_access(handle
, bh
);
4174 memcpy(bh
->b_data
+offset
, data
, len
);
4175 flush_dcache_page(bh
->b_page
);
4178 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
4180 /* Always do at least ordered writes for quotas */
4181 err
= ext4_jbd2_file_inode(handle
, inode
);
4182 mark_buffer_dirty(bh
);
4187 mutex_unlock(&inode
->i_mutex
);
4190 if (inode
->i_size
< off
+ len
) {
4191 i_size_write(inode
, off
+ len
);
4192 EXT4_I(inode
)->i_disksize
= inode
->i_size
;
4194 inode
->i_mtime
= inode
->i_ctime
= CURRENT_TIME
;
4195 ext4_mark_inode_dirty(handle
, inode
);
4196 mutex_unlock(&inode
->i_mutex
);
4202 static int ext4_get_sb(struct file_system_type
*fs_type
, int flags
,
4203 const char *dev_name
, void *data
, struct vfsmount
*mnt
)
4205 return get_sb_bdev(fs_type
, flags
, dev_name
, data
, ext4_fill_super
,mnt
);
4208 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4209 static struct file_system_type ext2_fs_type
= {
4210 .owner
= THIS_MODULE
,
4212 .get_sb
= ext4_get_sb
,
4213 .kill_sb
= kill_block_super
,
4214 .fs_flags
= FS_REQUIRES_DEV
,
4217 static inline void register_as_ext2(void)
4219 int err
= register_filesystem(&ext2_fs_type
);
4222 "EXT4-fs: Unable to register as ext2 (%d)\n", err
);
4225 static inline void unregister_as_ext2(void)
4227 unregister_filesystem(&ext2_fs_type
);
4229 MODULE_ALIAS("ext2");
4231 static inline void register_as_ext2(void) { }
4232 static inline void unregister_as_ext2(void) { }
4235 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4236 static inline void register_as_ext3(void)
4238 int err
= register_filesystem(&ext3_fs_type
);
4241 "EXT4-fs: Unable to register as ext3 (%d)\n", err
);
4244 static inline void unregister_as_ext3(void)
4246 unregister_filesystem(&ext3_fs_type
);
4248 MODULE_ALIAS("ext3");
4250 static inline void register_as_ext3(void) { }
4251 static inline void unregister_as_ext3(void) { }
4254 static struct file_system_type ext4_fs_type
= {
4255 .owner
= THIS_MODULE
,
4257 .get_sb
= ext4_get_sb
,
4258 .kill_sb
= kill_block_super
,
4259 .fs_flags
= FS_REQUIRES_DEV
,
4262 static int __init
init_ext4_fs(void)
4266 ext4_check_flag_values();
4267 err
= init_ext4_system_zone();
4270 ext4_kset
= kset_create_and_add("ext4", NULL
, fs_kobj
);
4273 ext4_proc_root
= proc_mkdir("fs/ext4", NULL
);
4274 err
= init_ext4_mballoc();
4278 err
= init_ext4_xattr();
4281 err
= init_inodecache();
4286 err
= register_filesystem(&ext4_fs_type
);
4291 unregister_as_ext2();
4292 unregister_as_ext3();
4293 destroy_inodecache();
4297 exit_ext4_mballoc();
4299 remove_proc_entry("fs/ext4", NULL
);
4300 kset_unregister(ext4_kset
);
4302 exit_ext4_system_zone();
4306 static void __exit
exit_ext4_fs(void)
4308 unregister_as_ext2();
4309 unregister_as_ext3();
4310 unregister_filesystem(&ext4_fs_type
);
4311 destroy_inodecache();
4313 exit_ext4_mballoc();
4314 remove_proc_entry("fs/ext4", NULL
);
4315 kset_unregister(ext4_kset
);
4316 exit_ext4_system_zone();
4319 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4320 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4321 MODULE_LICENSE("GPL");
4322 module_init(init_ext4_fs
)
4323 module_exit(exit_ext4_fs
)